Issuu

Production and Publication Research & Training, Dept. of Radiology & Nuclear Medicine, Erasmus MC

Scientific Report 2025

COLOPHON

Editor Kelly Zoutman Assistant Editors Lieke Visser Semanur Em

Printing GROENPRINT Aristotelesstraat 20 3076 BD Rotterdam The Netherlands For this scientific report, GROENPRINT and Trees for All plant several trees to restore the tropical rainforest Visiting Address Department of Radiology & Nuclear Medicine Erasmus MC Dr. Molewaterplein 40 3015 GD Rotterdam The Netherlands Telephone: +31 10 703 5372 research.radiology@erasmusmc.nl Post Address Department of Radiology & Nuclear Medicine Erasmus MC P.O. Box 2040 3000 CA Rotterdam The Netherlands Website Radiology & Nuclear Medicine – Department – Erasmus MC Cover photo Installation of the GE HealthCare Omni 128cm Total Body PET/CT

department of radiology & nuclear medicine

Design & Photography Lisa Weijman Steven Ensering Frank van der Panne Vincent Blinde

2025

Scientific Report department of radiology & nuclear medicine

2025

Scientific Report department of radiology & nuclear medicine

CONTENTS Preface Highlights 2025 Convergence Research Staff Research Support Imaging Facilities

5 6 14 18 22 28

BIOMEDICAL IMAGE ACQUISITION & ANALYSIS

32 34

Magnetic Resonance Physics in Medicine Juan A Hernández Tamames, PhD

Physics in CT Technology Marcel van Straten, PhD

AI in Medical Image Analysis Marleen de Bruijne, PhD

Image Guidance in Interventions and Therapy Theo van Walsum, PhD

Computational Population Biology

50 58

Gennady Roshchupkin, PhD

Applied Medical Image Analysis

Stefan Klein, PhD

Quantitative MRI Reconstruction Dirk Poot, PhD

NeuroImage Analysis & Machine Learning Esther E Bron, PhD

Musculoskeletal Image Analysis

Jukka Hirvasniemi, PhD

Artificial Intelligence for Integrated Diagnostics Martijn Starmans, PhD

Eye Image Analysis

88 96

Luisa Sanchez Brea, PhD & Danilo Andrade De Jesus, PhD

MOLECULAR IMAGING & THERAPY

102

Radiobiology of Radionuclide Therapy

104

Julie Nonnekens, PhD

112

Radiotracer Interactions Simone Dalm, PhD

Translational Nuclear Medicine

118

Frederik A Verburg, MD, PhD

Clinical Nuclear Medicine: Imaging and Therapy in Oncology Astrid van der Veldt, MD, PhD & Tessa Brabander, MD, PhD

124

Radiopharmaceutical Chemistry

130

Yann Seimbille, PhD

Theranostics of CNS and H&N diseases

138

Sophie Veldhuijzen van Zanten, MD, PhD

CLINICAL IMAGING

144 146

Imaging in Neurovascular Disease Aad van der Lugt, MD, PhD

152

Neuroradiology Marion Smits, MD, PhD

Bench-to-Bedside MR Imaging Biomarkers

158

Esther AH Warnert, ir, PhD

Cardiac Imaging

164

Ricardo Budde, MD, PhD & Alexander Hirsch, MD, PhD

Abdominal Imaging

172

Ivo G Schoots, MD, PhD

Advanced Musculoskeletal Imaging Research Erasmus MC (ADMIRE) Edwin HG Oei, MD, PhD

Interventional Radiology Research

180 190

Aad van der Lugt, MD, PhD

Advanced Thoracic Imaging

196

Daan Caudri, MD, PhD & Pierluigi Ciet, MD, PhD

IMAGING IN HEALTH SCIENCES

202 204

Population Imaging Meike W Vernooij, MD, PhD

Clinical Epidemiology of Neurological Disease Frank J Wolters, MD, PhD

Imaging of Arteriosclerosis: From Population to Clinical Practice Daniel Bos, MD, PhD

Value-based imaging

212 218 224

Jacob J Visser, MD, PhD

Pediatric Population Neuroimaging

230

Ryan Muetzel, PhD

238

INPUT & OUTPUT

240 242 244 246

Conference Contributions 2025 Publications 2025 Index Social Activities 2025

PREFACE

cientific progress is never the result of individual achievement alone. It emerges from curiosity, collaboration, perseverance, and a shared ambition to improve the lives of patients. When I reflect on the past year, I am above all struck by the dedication, creativity, and talent of the many colleagues who contribute every day to the scientific mission of the Department of Radiology & Nuclear Medicine at Erasmus MC.

Medical Image Analysis and Theo van Walsum Full Professor of Image Guidance for Interventions and Therapies. Pierluigi Ciet was appointed Associate Professor in recognition of his work on advanced pediatric thoracic imaging using CT and MRI. Simone Dalm was appointed Assistant Professor, focusing on the identification of tumor biomarkers and the development of novel radiotracers for targeted radionuclide theranostics. Esther Bron was appointed Assistant Professor to further develop innovative artificial intelligence methods for image analysis and outcome prediction in neurological and cardiovascular diseases. Frank Wolters was appointed Associate Professor in Clinical Epidemiology of Neurological Disease, while Roy Dwarkasing and Maarten Thomeer were appointed Assistant Professors in liver imaging and abdominal imaging, respectively. Their appointments reflect both individual excellence and the strength of our department’s scientific environment.

Our department thrives at the intersection of technology, clinical care, and scientific discovery. Rapid advances in imaging technologies, artificial intelligence, image-guided interventions, molecular imaging, population imaging continue to transform our field. Yet while technology is evolving at an unprecedented pace, our purpose remains unchanged: to better understand disease, improve diagnosis and treatment, and ultimately provide better care for our patients. This report highlights the quality and breadth of the research conducted within our department. From methodological innovations to large-scale clinical and population-based studies, our researchers continue to push the boundaries of imaging science. I am particularly proud of the collaborative nature of our work. Increasingly, important research questions can only be addressed by bringing together expertise from different disciplines, departments, institutions, and countries. The achievements described in this report reflect not only scientific excellence, but also the openness and collaborative spirit that characterize our department.

Collaboration remains essential to our success. During the past year, we further strengthened our partnership with the Department of Radiology at the University of Wisconsin. A twoday scientific symposium brought together researchers from both institutions to exchange ideas, share expertise, and explore new opportunities for collaboration. Such initiatives underscore the importance of international partnerships in addressing complex scientific and clinical challenges. None of the achievements described in this report would have been possible without the collective efforts of many people. I would like to express my gratitude to all researchers, support staff, clinicians, collaborators, funding agencies, industry partners, and, most importantly, the patients who participate in our studies and make our research possible. Their trust, commitment, and contributions form the foundation of everything we do.

One of the greatest privileges of leading this department is witnessing the development of the next generation of scientists. Their enthusiasm, curiosity, and ambition drive innovation and give me confidence that the future of imaging research is in excellent hands.

As we look ahead, healthcare faces significant challenges, but also unprecedented opportunities. New imaging technologies, advanced computational methods, and expanding international collaborations are opening doors that were unimaginable only a few years ago. By continuing to foster a culture of scientific excellence and collaboration, I am confident that our department will remain at the forefront of imaging research and continue to contribute to better healthcare and improved outcomes for patients.

During the past year, fifteen PhD candidates successfully defended their theses. While this may appear to be a routine academic milestone, every dissertation represents years of dedication, resilience, and hard work. I am proud of what these young researchers have accomplished and grateful for their contributions to our department. Whether they continue their careers in academia, healthcare, or industry, I am confident they will make a meaningful impact through the knowledge and skills they have developed here.

I hope this annual report not only reflects our scientific accomplishments, but also conveys the passion, ambition, and collective spirit that define our department.

Supporting talent remains one of our strategic priorities. The talent development program that was introduced in 2023 is now entering its final year, and next year we will begin a new cycle of talent reviews. Through this initiative, we aim to identify future leaders and provide them with the guidance, opportunities, and support needed to take the next step in their academic careers. It is rewarding to see how this program has already contributed to the advancement of several outstanding colleagues.

Aad van der Lugt, MD, PhD Professor and Chair Department of Radiology & Nuclear Medicine Erasmus MC, University Medical Center Rotterdam

The past year also marked important career milestones for a number of our researchers. Within the Biomedical Imaging Group, Stefan Klein was appointed Full Professor of Applied

HIGHLIGHTS 2025

Appointments Stefan Klein was appointed as Full Professor in Applied Medical Image Analysis.

(ECR) 2026 and was also appointed Neuroradiology Programme Planning Chair for the ECR 2027.

Theo van Walsum was appointed as Full Professor in Image Guidance for Interventions and Therapies.

Meike Vernooij was appointed as Vice-President for the ESNR annual meeting, to be held in London in October 2026.

Esther Bron was appointed as Associate Professor. Edwin Oei became Vice-President of the International Society of Osteoarthritis Imaging.

Simone Dalm was appointed as Associate Professor. Pierluigi Ciet was appointed as Associate Professor.

Edwin Oei became Vice-Chair of the Research Committee of the European Society of Musculoskeletal Radiology. He also served on the Musculoskeletal Scientific Programme Committee of the RSNA and the AI Taskforce of the European Society of Musculoskeletal Radiology.

Frank Wolters was appointed as Associate Professor. Roy Dwarkasing was appointed as Assistant Professor. Maarten Thomeer was appointed as Assistant Professor.

David Hanff became President of the Musculoskeletal Radiology section of the Dutch Society of Radiology.

Esther Warnert became Chair of the Glioma MR Imaging Working Group of the European Society of Magnetic Resonance in Medicine and Biology.

David Hanff became a member of the Tumor Subcommittee of the European Society of Musculoskeletal Radiology and joined the Congress Committee of the Netherlands Association of Sports Medicine.

Esther Warnert became President of the BeNeLux Chapter of the International Society for Magnetic Resonance in Medicine.

Simone Dalm became Associate Editor of the British Journal of Radiology.

Esther Warnert became Chair of the Network for Ambitious Women at Erasmus MC.

Marleen de Bruijne was guest editor for the Special Issue on “Advancements in Foundation Models for Medical Imaging” from IEEE Transactions on Medical Imaging.

Rianne van der Heijden was appointed as Adjunct Clinical Assistant Professor at the Department of Radiology, University of Wisconsin–Madison, USA.

Simone Dalm was guest editor for EJNMMI reports on “the Role of Radiopharmaceutical Agents in Cancer Theranostics”.

Rianne van der Heijden chaired the Erasmus MC–University of Wisconsin–Madison symposium.

Meike Vernooij joined the international advisory board of the Lancet Medical Imaging & Theranostics.

Rianne van der Heijden co-chaired the International Society for Magnetic Resonance in Medicine musculoskeletal workshop.

Contribution to Guidelines

Daniel Bos was appointed as Secretary of the Board for Review and Assessment of the Netherlands Epidemiological Society.

Marion Smits was involved in the Update of the RANO Working Group and EANO recommendations for the clinical use of PET imaging in gliomas. and the ESTRO/ EANO recommendation on re-irradiation of glioblastoma.

Meike Vernooij acted as Neuroradiology Programme Planning Chair for the European Congress of Radiology

Martijn Starmans co-led the development of the FUTURE-AI guideline for trustworthy and deployable AI in healthcare.

Meike Vernooij was interviewed by RTV Rijnmond on the occasion of 35 years of ERGO, the Rotterdam Study.

Roy Dwarkasing contributed to ongoing international efforts to further refine and improve the LI-RADS classification system for liver imaging and hepatocellular carcinoma diagnosis.

Chiara Bruck, Inge de Kok and Frank Wolters organized a public session on dementia prevention at the National Dementia Conference, organized by the Ministry of Health. Wouter Schallig’s research in the Motion Biomechanics & Imaging Lab was featured in the national and regional public media.

François Willemssen contributed to two Dutch guidelines on diagnostic abdominal imaging for hepatocellular carcinoma and cholangiocarcinoma.

Daniel Bos delivered two layman’s presentations on the role of intracranial vascular disease in dementia for Alzheimer Nederland, and Cure Alzheimer’s Fund.

Maarten Thomeer contributed to the Dutch guidelines for ovarian, endometrial and cervical carcinoma. Ricardo Budde led the writing group of the multi-society consensus document on CT imaging of prosthetic heart valves.

Awards Rianne van der Heijden’s article “Platelet-Rich Plasma Therapy for Patellar Tendinopathy: A Randomized Controlled Trial Correlating Clinical Outcomes and Novel Imaging Biomarkers” was selected as an Editor’s Pick for the Best of Radiology Advances 2024 Top 5 papers.

Ivo Schoots is Co-Chair of the international PI-RADS Steering Committee on prostate MR imaging, is a full panel member of the European Association of Urology Prostate Cancer Guidelines and is advisor to the Dutch Cancer Foundation Committee on development and implementation.

Gennady Roshchupkin won the Frans van der Helm Award for Best Young Researcher in the field of biomedical engineering.

Astrid van der Veldt was chair of the working group on the new Dutch Melanoma Guideline.

Priciana Paraiso and Le Li received Travel Awards from the Society of Radiopharmaceutical Sciences to present their work at the International Symposium of Radiopharmaceutical Sciences 2025.

Alexander Hirsch contributed to the consensus statement on multi-modality cardiac imaging in hypertrophic cardiomyopathy of the European Association of Cardiovascular Imaging of the European Society of Cardiology.

Circe van der Heide won the Young Investigator Award of the European Association of Nuclear Medicine.

Mariëlle Olsthoorn contributed to the new national guideline on acute muscle injury of the lower extremity in athletes.

Jie Deng was awarded a Collaborative Exchange Grant by the Orthopaedic Research Society. With this exchange, Jie prepared her transition to a postdoctoral position at the University of Wisconsin–Madison.

Societal Impact Patrick Tang contributed to the public understanding of research as one of the Royal Netherlands Academy of Arts and Sciences (KNAW) ‘Faces of Science’.

Patrick Tang won first prize in the General Audience Competition at the annual meeting of the European Society for Magnetic Resonance in Medicine and Biology in October 2025.

AuntMinnie published an article on pain imaging research performed by the group of Edwin Oei.

Patrick Tang was nominated for the Young Investigator Award at the annual meeting of Glioma MR Imaging 2.0 (GliMR) in October 2025. Tareq Abdel Alim won first prize for best research presentation with “Predictive Modelling of Postoperative Cranial Shapes using Deformation Vector Fields” at the European Society for Pediatric Neurosurgery 2025 in London, UK, on 14 May 2025.

Frank Wolters, Muhammed Sambou and Sanne van Kuijk participated in a panel discussion and public education activities on science and prevention during festivities with Rotterdam Study participants, held in honor of the 35th anniversary of the cohort.

Tareq Abdel Alim’s article ‘Quantifying dysmorphologies of the neurocranium using artificial neural networks’ was named top cited paper by Journal of Anatomy.

Wietske Bastiaansen received the Best Poster Award at the kick-off event of the Erasmus MC AI Accelerator for her research on prenatal image analysis.

Frank Wolters co-organized the VasCog international conference in Southampton, UK, as part of the scientific and executive committees of VasCog, the International Society of Vascular Behavioural and Cognitive Disorders. Wouter Schallig co-organized a session entitled “Towards a community-endorsed research environment to enable standardised data sharing in the Human Movement Sciences” during the International Society of Biomechanics congress in Stockholm from 27–31 July 2025. Wouter Schallig was the lead organizer of the annual congress of the Society for Movement Analysis Laboratories in the Low Lands, held from 27–28 November 2025 in Rotterdam.

Ilanah Pruis received the prestigious Alavi-Mandell Award from the Society of Nuclear Medicine and Molecular Imaging and the Journal of Nuclear Medicine for the publication “[18F]FET PET/MRI: An Accurate Technique for Detection of Small Functional Pituitary Tumors” in April 2025.

Rianne van der Heijden, as lead organizer, and Edwin Oei cohosted the inaugural Erasmus MC–University of Wisconsin– Madison symposium from 12–14 March 2025. As part of the strategic partnership between our department and the Department of Radiology of the University of Wisconsin– Madison, the symposium focused on establishing and strengthening collaborations in musculoskeletal imaging, nuclear medicine, image analysis and physics.

Ilanah Pruis received the EANM Young Investigator Award at the EANM Young Investigator Meeting 2025 in Vienna, Austria, for the ACT on PitNET study in June 2025.

Renske Gahrmann was co-organizer of the Rotterdam Working Group Head and Neck Tumors scientific symposium in Rotterdam, the Netherlands.

Mariangela Sabatella won the Best Poster Presentation Prize at the International Workshop on Radiobiology of Molecular Radiotherapy in April 2025.

Ivo Schoots co-organized the RSNA/ECR 2024–2025 and 2025–2026 joint transatlantic sessions on prostate cancer imaging at the Radiological Society of North America and the European Congress of Radiology.

Nina Becx won the Best Oral Presentation Prize at the International Workshop on Radiobiology of Molecular Radiotherapy in April 2025.

Stefan Klein and Esther Bron organized a HealthRI Imaging Community meeting on the use of Secure Processing Environments, an important technique for safe data sharing as suggested by the European Health Data Space regulation.

Pleun Engbers won the Klaas Breur Travel Grant for best presentation at the annual meeting of the Netherlands Society for Radiobiology in November 2025.

Pierluigi Ciet organized the European Course of Pediatric Radiology Module II on neck, chest and cardiovascular imaging in Rotterdam.

Kaouther Mouheb and Myrthe van Haaften were awarded TAP-Connect grants from the TAP-Dementia consortium to explore how cross-cohort heterogeneity impacts the performance of AI models trained for dementia diagnosis. Frank te Nijenhuis received a DAAD (German Academic Exchange Service) grant to visit the Charite Lab for Artificial Intelligence in Medicine in Berlin.

Meetings Frank te Nijenhuis and Theo van Walsum co-organized the 5th SWITCH Workshop at the MICCAI conference.

Dr Hong Liu, MD, PhD, from the Clinical Medical School, Lanzhou University, Lanzhou, China, was a visiting research fellow at our department. She conducted collaborative research under the supervision of Roy Dwarkasing, Kay Pieterman and Piotr Wielopolski.

Frank te Nijenhuis, Xi Li, Ellen Hu, Peter van Hulst, Govert van der Gun and Theo van Walsum organized a workshop on stroke care at the Rotterdam Stroke Service Symposium on 4 November 2025.

scientific report 2025 | HIGHLIGHTS

In June 2025, representatives of the Taiwan Smart Healthcare Association visited Erasmus MC. Discussions focused on strengthening scientific collaboration and clinical cooperation between Taiwan and Erasmus MC. These discussions resulted in a first joint research project on AI-assisted diagnosis of small pancreatic cancers, funded by the Taiwan Smart Healthcare Association. Martijn Starmans continued his work as co-lead of the MICCAI Open Data initiative, which he co-founded in 2024 to facilitate and promote the sharing of medical imaging datasets. In this context, he was part of the MICCAI 2025 organization. As part of the track, Kaouther Mouheb co-organized a main event session and a special issue, including the publication of 13 new datasets. For this work, the initiative received a MICCAI Health Equity Award. Within Theo van Walsum’s group, there were four contributions to the IPCAI 2025 conference in Berlin, including contributions from four MSc students. Sophie Veldhuijzen van Zanten was invited to deliver a keynote lecture entitled “Quo Vadis Nuclear Medicine?” at the 38th Annual Congress of the European Association of Nuclear Medicine in Barcelona, Spain.

Grants 2025 Personal Grants / Fellowships Dutch Research Council VIDI grant

Frank Wolters

Gepersonaliseerde diagnostiek en behandeling van stille herseninfarcten (SHINOBI-II)

Dutch Research Council VIDI grant

Ryan Muetzel

Understanding the antecedents and consequences of substance use in youth

Collaborative Exchange Grant

Jie Deng

Rehabilitation strategies by integrating advanced imaging modalities with an innovative wearable device to assess patellar tendon loading.

European Association of Nuclear Medicine Young Investigator Grant

Ilanah Pruis

The ACT on PitNET Study – Accelerating European-Wide Clinical Translation of AI-enhanced Amino Acid-Based Radionuclide Diagnostics, Treatment Planning and Disease Monitoring for Patients with Pituitary Neuroendocrine Tumors

Ministry of Education, Culture & Science Starting Grant

Julia Neitzel

Brain Health

Ministry of Education, Culture & Science Starting Grant

Martijn Starmans

RadPath‑RI: Research infrastructure to join forces of radiology and pathology through Artificial Intelligence for Integrated Diagnostics (AIID)

Erasmus MC Young Investigator Grant

Wouter Schallig

Precision diagnostics of joint function by converging biomechanics and imaging

Royal Netherlands Academy of Arts and Sciences, Van Leersum Grant

Laura Kemper

Improving clinical applicability of APTw-CEST MRI for glioblastoma treatment response evaluation

Dutch Research Council Cost-Effectiveness Research

Daniel Bos

Clinical Impact of the IMPROVE model for patient selection for carotid endarterectomy

Dutch Research Council Off Road Grant

Joana Campeiro

Exploiting aberrant proteoglycan expression profiles for targeted therapy of cancer

Dutch Research Council XS Open Competition

Simone Dalm & Esther Warnert

Characterizing prostate cancer lesions that are resistant to PSMA-targeted radionuclide therapy for the development of a more personalized approach

Health Holland

Aad van der Lugt & Theo van Walsum

PP CONTRAST: Public-Private COllaboration for Novel Technologies for Acute Stroke

National Grants

Health Holland

Gennady Roshchupkin

AI based personal hormone profiling for the protection of mother and child’.

Health Holland

Luisa Brea Sanchez & Danilo Andrade de Jesus

Improving OCTA image analysis to enhance ophthalmic and systemic vascular CARE (OCTA-CARE)

Health Holland

Theo van Walsum & Kay Pieterman

See-through-vision for surgeons: Tumor localization and visualization using magnetic seed tracking and augmented reality

Oncode

Julie Nonnekens

Massive single-cell DNA damage profiling to uncover vulnerabilities of cells treated with alpha and beta radionuclide therapy

Medtech Thematic Technology Transfer Grant

Erik de Blois

Xtender: Radiopharmaceuticals shelf-life extender

Dutch health insurers

Jan Jaap Visser

AI for Imaging evaluation project

European Innovation Council Accelerator Initative

Yann Seimbille

eFAP Drugs for Imaging and Treatment of Cancer

EU Horizon Europe

Martijn Starmans, Stefan Klein & et al.

A Pan‑African Research Infrastructure for Collaborative Biomedical Imaging and Artificial Intelligence (AFRICAI‑RI)

EU Innovative Health Initiative (IHI)

Edwin Oei

Patient Relevant Osteoarthritis endpoints using Big data Evaluation (PROBE)

EU4Health

Sophie Veldhuijzen van Zanten, Marcel van Straten, Pierluigi Ciet & Marjolein Dremmen

Radiation, Health, Safety and Quality for Youth: A Comprehensive Approach to Justification, Optimisation, and Education (RHYTHM)

Digital Europe Programme (DIGITAL)

Jan-Jaap Visser, Ilva van Houwelingen, Stefan Klein & Martijn Starmans

SHAIPED - Supporting Health Data Access Bodies to establish AI pathways enabling Deployment of AI as medical device tools

International Grants

Digital Europe Programme (DIGITAL)

Jifke Veenland

Sustainable Healthcare with Digital Health Data Competence (SUSA)

Euratom, European Partnership for Radiation Protection Research

Simone Dalm

Framework for Outcomes of Radionuclide Radiation on Cellular Environments (FORRCE)

Swiss National Science Foundation

Sophie Veldhuijzen van Zanten

Use of [18F]FET PET-MRI to improve detection of pituitary adenomas in Cushing’s disease a prospective diagnostic study

scientific report 2025 | HIGHLIGHTS

Charitable Organisations Alzheimer Nederland Cross-border Grant

Julia Neitzel

Early plasma biomarker signature of Alzheimer in the general population

Brain Tumour Foundation

Sophie Veldhuijzen van Zanten

Intra-arterial targeted radionuclide therapy for glioma (ALPINE)

Brightfocus Foundation Grant

Daniel Bos

Unraveling Vascular Pathways in Dementia through Integrated Computational Fluid Dynamics

Dutch Cancer Foundation

Sophie Veldhuijzen van Zanten & Tessa Brabander

Intra-arterial treatment of meningioma with alpha emitting radionuclide therapy

Dutch Heart Foundation

Daniel Bos & Ricardo Budde

ATHERONETH - Defining strategies to dissect the diversity in atherosclerotic disease and stratify patients for future treatment

Stichting Semmy

Sophie Veldhuijzen van Zanten

Theranostics for CNS tumours

Convergence: Healthy Start Sprint

Ryan Muetzel

Touching Minds

Erasmus MC EraSupport

Martijn Starmans

Detecting Complex Regional Pain Syndrome and quantifying treatment success using a trustworthy artificial intelligence model based on video thermography

Erasmus MC EraSupport

Xinyi Wan, Martijn Starmans, Stefan Klein, Edwin Oei & JanJaap Visser

Trustworthy and explainable AI models for differentiating benign and malignant bone tumors on radiological imaging

Erasmus MC EraSupport

Eline van Lange

Detecting Complex Regional Pain Syndrome and quantifying treatment success using a trustworthy artificial intelligence model based on video thermography

Erasmus MC Trustfonds

Esther Warnert

Validation of liver oxygen MRI

Erasmus MC Trustfonds

Wouter Schallig

Bike Better: Optimize healthy joint loading during cycling

Institutional Grants

Investigator initiated industry sponsored grants GE HealthCare

Erik Verburg

Evaluation of the novel GE Healthcare Omni 128 cm LAFOV PET/CT

RayzeBio

Julie Nonnekens

Investigating incomplete responses to actinium-225-based TRT: mechanisms and cellular heterogeneity

Varian

Ivo Schoots

Organ-sparing marker-less CBCT-guided stereotactic adaptive radiotherapy for primary non-metastasized renal tumors

New Facilities

Philips Medical Systems Epiq Elite One of our Ultrasound machines, used in interventional radiology, was replaced in 2025. Erasmus MC has a contract with Philips Medical Systems for ultrasound equipment for the whole hospital. The new system for interventional radiology is a Epiq Elite, the high end system in the Philips portfolio.

Siemens CIOS Spin 3D C-arms Under a three-year lease agreement, two Siemens CIOS Spin 3D C-arms have been procured. This equipment is intended for use in operating rooms, particularly for complex spinal surgery, and is supported by software from Globus Medical. During procedures, the system enables the acquisition of high-quality three-dimensional images of the spine, providing the surgeon with enhanced visualization compared to previous methods. It is anticipated that this will lead to reduced operating times, improved surgical outcomes, and a decrease in the number of post-operative CT scans required.

GE Healthcare OMNI Legend 32 In the first months of 2025 we installed a new PET-CT scanner, the GE Healthcare OMNI Legend 32. It replaces on of our old PET-CT’s. With it’s 32 cm PET detector it makes it possible to scan patients faster than before.In 2026 the OMNI Legend 32 will be upgraded to a OMNI Legend 64, with a detector range of 64 cm. In November GE Healthcare started with the installation of the OMNI Legend 128. It was the second installation of this new type of scanner in the world. The new system has an axial field of view of 128 cm, enabling whole body PET-CT imaging, capturing more anatomy at once and supporting advanced applications. The installation was completed in December; ready for scanning the first patients in January 2026.

scientific report 2025 | HIGHLIGHTS

Atle® 180 Since the summer of 2025, the CT unit has implemented the Atle® 180 as a patient transfer support device. The Atle® 180 is designed to reduce the physical workload of healthcare professionals while enabling efficient patient transfers with a particular focus on radiology departments. The device allows for lateral patient transfer at the push of a button and can independently move patients weighing up to 180 kg. For patients exceeding this weight, it significantly reduces the amount of manual effort required, thereby limiting physical strain on healthcare professionals, especially on the lower back. In radiology departments, healthcare professionals frequently perform manual transfers when moving patients between beds and imaging or procedure tables. The frequency of these transfers has increased because of ongoing technological advancements in imaging equipment.

Carestream DRX Revolution Plus For the conventional modality, two mobile digital X-ray units, the Carestream DRX Revolution Plus, have been acquired. These units function as fully digital X-ray rooms on wheels and are specifically designed for use in departments such as intensive care units (ICUs), emergency departments (EDs), and operating rooms. Equipped with a wireless DRX Plus detector, the system rapidly produces highquality images directly at the patient’s bedside. These images are immediately transferred to the digital archive, ensuring they are instantly available for review and reporting.

CONVERGENCE

Erasmus MC, TU Delft, and the EUR aim to become a global leader in Health & Technology Convergence The urgent and complex social issues of our time demand convergence; looking beyond the boundaries of institutions and disciplines to arrive at new perspectives and solutions. For this reason, TU Delft, Erasmus MC and Erasmus University Rotterdam have joined forces in the Convergence. Convergence is built around five programs on specific societal challenges: Resilient Delta; Health & Technology; AI, Data & Digitalisation; Pandemic & Disaster Preparedness and Health Start. At the three institutions, over 300 principal investigators already cover the required range of disciplines. In research and education, their strengths, knowledge and methods to tackle problems related to climate change, sustainability, pressure on the healthcare system, urbanization and digitalization will be combined. The delta around Rotterdam and Delft – with its companies, harbors, rivers and diverse population – is a ‘living lab’. Together with public and private partners, a robust ecosystem for research and innovation will be build.

Figure: Ten Flagships started in 2022.

In 2022 ten Convergence Flagship programs have started to conduct research aimed at making healthcare future proof. The Flagships will receive a total of 20 million euros from the Convergence Health & Technology program. Collaboration with public and private parties adds 34 million to that. The investment in joint research is a solid step forward in the collaboration between the three institutions. Together they have what it takes to become a global leader in Health & Technology Convergence, shaping the future of health and healthcare in a transformative way. The convergence themes aim to respond to urgent challenges, like staffing and healthcare expenses, and where many healthcare need currently remain unmet, the flagships aim to transform our healthcare system to become more proactive, personalized, precise, participatory and labor friendly. At the same time, the convergence approach of integrating knowledge, expertise and methodologies from different domains will foster groundbreaking new scientific discoveries, pushing the state of the art. The department of Radiology & Nuclear Medicine has a lot of expertise and knowledge, has many technological devices, smart software programs and large image datasets. Image acquisition and interpretation requires in depth knowledge on medical physics. Image analysis requires engineering and programming skills. Today artificial intelligence contributes tremendously to image acquisition, analysis and interpretation. As imaging department we play an important exemplary role in the Health & Technology convergence and are involved in different flagships.

Brain Tumors The Brain Tumors Convergence program focuses on improving diagnosis and treatment of brain tumors through advanced imaging and data-driven approaches. Researchers combine expertise in MRI, image analysis, artificial intelligence, and clinical diagnostics. The program evolved from the flagship project Deep Medical Imaging of Structure, Physiology and Function and now contributes to broader cancer diagnostics initiatives, including theranostics, intraoperative imaging, and early cancer detection. Its strength lies in integrating clinical, technical, and social science perspectives while leveraging advanced facilities such as PET-MRI and proton therapy.

Erasmus University and Erasmus MC Collaboration on Advanced fMRI Facility The collaboration between Erasmus University Rotterdam (EUR) and Erasmus MC has resulted in the establishment of an advanced MRI facility equipped with functional MRI (fMRI) technology. The facility supports innovative research in neuroscience, psychology, and medical science and serves as a hub for current and future interdisciplinary studies.

Ongoing projects Erasmus MC

Musculoskeletal Imaging This initiative combines radiology, biomechanics, artificial intelligence, and orthopedics to improve the understanding and diagnosis of musculoskeletal disorders. Researchers develop quantitative imaging biomarkers from MRI and PET data and integrate these with biomechanical measurements. A major achievement is the establishment of the MOtionBiomechanics & Imaging (MOBI) Lab, a joint Erasmus MC–TU Delft facility that enables close collaboration between clinicians, engineers, imaging scientists, and physicists. The goal is to improve personalized diagnosis and treatment of musculoskeletal diseases.

Several projects are currently making use of the facility: 1.

Image-Guided Therapy The Image-Guided Therapy program aims to create smarter and more precise minimally invasive interventions. By combining surgical expertise, optical sensing technologies, augmented reality, artificial intelligence, and medical imaging, researchers are developing systems that provide surgeons with real-time information during procedures. Projects include innovations for stroke treatment, liver ablation, and navigation systems that improve procedural accuracy. The collaboration demonstrates how engineering and clinical expertise can jointly advance next-generation surgical technologies.

Theranostics Theranostics combines diagnostic imaging and targeted therapy using radiopharmaceuticals. The chapter highlights collaboration between Erasmus MC and TU Delft in developing new radioactive tracers, imaging methods, and nuclear microscopy technologies. Researchers are working on alpha-labelled radiopharmaceuticals and advanced imaging systems that improve both detection and treatment of disease. The partnership benefits from TU Delft’s expertise in isotope production and technology development, while Erasmus MC contributes clinical and biomedical applications. This area is presented as an example of how convergence can accelerate innovation in precision medicine.

Generation R (GenR) investigates genetic and environmental factors influencing health and disease across generations. A new cohort will be recruited in 2025. PANDA studies neurodevelopmental processes and their relationship with mental health using advanced neuroimaging techniques. OPPER focuses on severe postpartum mood disorders, including depression, mania, and psychosis. Through blood sampling and neuroimaging, the study aims to identify biomarkers, understand disease mechanisms, and predict long-term outcomes, while distinguishing postpartum-specific disorders from classical bipolar disorders. BRIDGE is developing brain growth charts for children and adolescents aged 6–20 using more than 7,000 brain scans from the Generation R cohort. In collaboration with cohorts in the United States and China, the project aims to improve detection of atypical brain development and support clinical decision-making in pediatric neuroradiology.

EUR Research Initiatives Several EUR-led studies have also commenced: 1.

Braintime was completed in 2024 and marked the first collaboration between the SYNC Lab and the Department of Radiology. The study examined neural correlates of well-being in young adults and resulted in the first joint fMRI publication. Growing Up Together in Society (GUTS) is a 10year longitudinal study investigating the conditions for successful development in a complex society. Through multiple imaging waves, it examines brain development, self-regulation, trust, and behavioral outcomes in adolescents from diverse backgrounds. SocCRED explores how social credit scores affect neural activity during trust-based decisions and whether such systems reinforce or reduce social biases. The findings may inform policy and future research on social decision-making.

GMG (Neural Prediction of Multiattribute Giving) investigates the neural mechanisms underlying charitable donations by examining brain responses to real-life donation appeals and how evaluations change over time.

Convergence Labs A vision for future research The success of these studies highlights the potential of the MRI facility to support a broad range of interdisciplinary research. To facilitate future projects, an fMRI Commission has been established, led by Eveline Crone, Maarten Boksem, Ryan Muetzel, Carolina Deurloo-Mendez Orellana, Monique de Waard, and Muhammet Sahan. The commission oversees the development, coordination, and optimal use of the facility and its resources.

State-of-the-art facility to assess joint motion Approximately a dozen biomechanical motion analysis laboratories exist in the Netherlands, but their accuracy is limited due to so-called soft tissue artifacts. The MOBI Lab is the first facility in the Benelux to combine biomechanical motion analysis with videoradiography (i.e. fluoroscopy). By acquiring 15 radiographs per second while a participant is moving, we can directly visualize and track bone motion, enabling highly accurate assessment of joint kinematics in unprecedented detail.

The Motion Biomechanics & Imaging (MOBI) laboratory The Motion Biomechanics & Imaging (MOBI) Lab is located within the Department of Radiology & Nuclear Medicine (Nd-546k). It is a collaboration with the Department of Orthopedics & Sports Medicine and TU Delft. Officially opened in October 2024, it is the first shared facility of Convergence Health & Technology.

Figure 2: Videoradiography images during walking with a lateral (2 times) and frontal view, and during a deep knee bend.

Current projects The initial, primary focus of lab is to advance the understanding, diagnosis, and treatment of osteoarthritis, one of the most prevalent joint disorders worldwide. The first two major research studies quantify joint loading during walking, running, and cycling in individuals with established osteoarthritis or at increased risk of developing the disease, such as those with obesity or traumatic knee injuries. These data are integrated with PET-MRI findings to provide a comprehensive assessment of joint health.

A vision for future research and clinical application The MOBI Lab can become a foundation for future collaborative clinical research, with applications extending beyond osteoarthritis to other joints and musculoskeletal disorders. In addition, we strive to establish this technology as a clinically valuable tool within the next five years.

Figure 1: Overview of the Motion Biomechanics and Imaging Lab and a zoomed in picture when the system is in use.

scientific report 2025 | CONVERGENCE

RESEARCH STAFF Rianne van der Heijden, MD, PhD Roy Dwarkasing, MD, PhD Ryan Muetzel, PhD Sophie Veldhuijzen van Zanten, MD, PhD Tessa Brabander, MD, PhD

Post-Docs & Junior Researchers Anke van der Eerden, MD, PhD Annick Weustink, MD, PhD Arlette Odink, MD, PhD Eline Vinke, PhD Erik Vegt, MD, PhD Fariba Tohidinezhad, PhD Giulia Tamborino, PhD Guillermo Garrido Hernandez, PhD Hanyue Ma, PhD Hyunho Mo, PhD Ilanah Pruis, PhD Ilva Klomp, PhD Joana Campeiro, PhD Justine Perrin, PhD Kay Pieterman, MD, PhD Laura Nunez Gonzalez, PhD Lucie Arnaud, PhD Mariangela Sabatella, PhD Mark Konijnenberg, PhD Mark de Wolf, MD, PhD Maryana Handula, PhD Mohamed Benmahdjoub, PhD Renske Gahrmann, MD, PhD Rob van de Graaf, MD, PhD Ronald Booij, PhD Sandra Cornelissen, MD, PhD Tareq Abdel Alim, PhD Ties Mulders, MD, PhD Wouter Schallig, PhD

Full Professors Aad van der Lugt, MD, PhD Edwin Oei, MD, PhD Frederik Verburg, MD, PhD Juan Hernández Tamames, PhD Marion Smits, MD, PhD Marleen de Bruijne, PhD Meike Vernooij, MD, PhD Myriam Hunink, MD, PhD Ricardo Budde, MD, PhD Stefan Klein, PhD Theo van Walsum, PhD

Associate Professors Alexander Hirsch, MD, PhD Astrid van der Veldt, MD, PhD Daniel Bos, MD, PhD Esther Bron, PhD Frank Wolters, MD, PhD Frans Vos, PhD Ivo Schoots, MD, PhD Jifke Veenland, PhD Julie Nonnekens, PhD ius promovendi Pierluigi Ciet, MD, PhD Simone Dalm, PhD Stijn Koolen, MD, PhD Yann Seimbille, PhD

Assistant Professors Daan Caudri, MD, PhD Danilo Andrade de Jesus, PhD Dirk Poot, PhD Esther Warnert, PhD Erik de Blois, PhD Gennady Roshchupkin, PhD Gyula Kotek, MD, PhD Henri Vrooman, PhD Jacob Visser, MD, PhD Jukka Hirvasniemi, PhD Julia Neitzel, PhD Luisa Sánchez Brea, PhD Maarten Leening, MD, PhD Maarten Thomeer, MD, PhD Marcel van Straten, PhD Martijn Starmans, PhD

PhD Students Including partially appointed to department Abdullah Thabit, MSc, PhD 2025 Adnane Zerguit, MSc Adriana Falcao Neves, MSc Ahmad Alafandi, MD, MSc Aikaterini Tziotziou, MSc, PhD 2025 Alexander Wakker, MD, MSc Alireza Samadifardheris, MSc Amber Audhoe, MSc Ana Gebara Garcia de Lima, MSc Angelina Pieters, MD, MSc, PhD 2025 Anna Streiber, MSc Anna Lavrova, MD, MSc Anne Akerboom, MSc Arno van Hilten, MSc Asabi Leliveld, MSc

Bart-Jan Boverhof, MSc Bas Dille, MSc Bas Meyer Viol, MD, MSc Bianca Dijkstra, MSc Bina Tariq, MD, MSc Bo Li, MSc Boudewijn Willems, MSc Bram Roumen, MSc Brian Berghout, MSc Bridget Schoon, MD, MSc Brigit van Dijk, MD, MSc Britt van Dijk, MSc Camiel Box, MD, MSc Carolline Ntihabose, MSc Céline van de Braak, MSc Ching Khan, MD, MSc Chintan Chawda, MSc Christina Cretu Circe van der Heide, MSc Daniek van der Kaaij, MSc David Hanff, MD, MSc Desirée de Vreede, MD, MSc Diede Wijnbergen, MSc Dorottya Papp, MSc Douwe Spaanderman, MSc Duscka Kleijn, MSc Duygu Harmankaya, MD, MSc Dylan Chapeau, MSc Eefje Dalebout, MD, MSc Eline Hooijman, MSc, PhD 2025 Eline Zoetelief, MSc Elisa Maijoor, MSc Emanoel Sabidussi, MSc Erik Kemper, MSc Esther Droogers, MD, Msc Eva Bocharewicz, MD, MSc Federico Mollica, MSc Felipe Gama Franceschi Francois Willemssen, MD, MSc Frank te Nijenhuis, MD, MSc Frederik Hartmann, MSc Gerda Verduijn, MD, MSc Gigi Vissers, MSc Gonzalo Mosquera Rojas, MSc Guanqun Liu, MSc Hannelore Coerts, MSc Hazel Zonneveld, MD, MSc Huib Ruitenbeek, MSc, PhD 2025 Ieva Aliukonyte, MSc Ikenna Asuzu, MSc Ilaria Neri, MSc Ilaria Orsi, MSc Imren Ozdamar, MD, MSc Ingrid Bakker, MSc Jacqueline Claus, MD, MSc, PhD 2025 Jamie Verwey, MSc

Jarno Steenhorst, MSc Jasika Paramasamy, MSc Jessica de Jong, MD, MSc Jiaqi Tang, MSc Jie Deng, MD, MSc Jing Yu, MSc Jochem Wolfert, MSc Joep van de Sanden, MSc José Castillo Tovar, MSc, PhD 2025 Juancito van Leeuwen, MSc Judith van der Bie, MSc Julie Hamm, MSc Kaouther Mouheb, MSc Karen van der Werff, MSc Karst Adriaansens, MSc Karlijn de Joode, MD, MSc, PhD 2025 Karthik Prathaban, MSc Katrien Bracké, MD, MSc Krishnapriya Venugopal, MSc Laura Kemmeren, MSc Laura Kemper, MSc Laura Monteiro Rente Dias, MSc Laurens Topff, MD, MSc, PhD 2025 Lennard Wolff, MD, MSc Le Li, MSc Lisa Bokhout, MSc Luca Bontempi, MSc Lucas de Kruijff, MSc Mahlet Birhanu, MSc Mara Veenstra, MSc Marchella Zijta, MSc Matthew Marzetti, MSc Marguerite Faure, MD, MSc Mariana Silva Pereira Fialho de Piedade, MSc Marijn Mostert, MSc Marjolein Dremmen, MD, MSc Mark van den Dorpel,MD, MSc Marleen van den Heuvel, MD, MSc Mathijs Rosbergen, MSc Matthijs van der Sluijs, MD, MSc Matthijs van Dongen, MSc Maurits Blomjous, MSc Megan van de Veerdonk, MD, MSc Micha van der Lee, MD, MSc Mirthe Kamphuis, MSc Meedie Ali, MD, MSc Merel de Vries, MSc, PhD 2025 Myrthe van Haaften, MSc Nabil Loddin, MSc Natalia Oviedo Acosta, MSc Neslisah Seyrek, MD, MSc Niels Dur, MD, MSc Nicolai Herrmann, MSc Nikki Boodt, MD, MSc PhD 2025 Nina Becx, MSc Nina Overdevest, MSc

IT Architects and Research Software Engineers

Noemi Sgambelluri, MSc Patrick Tang, MSc Peter van Hulst, MSc Pinar Yilmaz, MD, MSc Pleun Engbers, MSc Pranali Raut, MSc Priciana Paraiso, PharMD Roisin McMorrow, MSc, PhD 2025 Rosemarijn van Paassen, MSc Ruben Niemantsverdriet, MSc Sanne Boeren, MD, MSc Shishuai Wang, MSc Simran Sharma, MD, MSc Sophie Derks, MD, MSc, PhD 2025 Stanislav Ivanov, MSc Sterre de Jonge, MSc Stijntje Dijk, MD, MSc PhD 2025 Subhradeep Kayal, MSc Swaaij Ling, MD, MSc Thom van der Laan, MSc Thuy Nguyen, MSc Tijmen van Zadelhoff, MD, MSc Tijmen de Wolf, MSc Tiny Cox, BSc Tyrillshall Damiana, MSc Wenjie Kang, MSc Xi Li, MSc Xiaoyu Zhang, MSc Xianjing Liu, MSc, PhD 2025 Xinyi Wan, MSc Zoë Keuning, MSc

Adriaan Versteeg Hakim Achterberg Ivan Bocharov Mahlet Birhanu Marcel Koek

Student Assistant MRI ERGO/GenR Anne-Sterre Schutter Amanda Ros Bjorn Schrijver Dagmar Mook Esra Hemmelder Fengli Bottema Fenne Bosman Gaia Hermans Gunnar Zwart Ismay van der Laar Issrae Affani Lara Uitdewilligen Levy Schimmel, Team Leader Lieke Bouvy Lucas de Groot Marije Timmermans Mats Abspoel Mehdi Badaoui Ouidad Oujjit Paula Rijs Alonso, Team Leader Semih Cakmaktepe Ties Muller

Technicians

Unit Research & Training

Amber Piet – Research Technician Corrina de Ridder - Biotechnican Debra Stuurman - Biotechnican Jan de Swart – Imaging Specialist Lisette de Kreij-de Bruin – Research Technician Marcel Dijkshoorn – Research Technologist CT Rob Verhagen – Research Technician

Monique de Waard – Director of Research & Training Lieke Visser – Secretary Research & Training

Unit Research & Training Imaging Trialbureau and Imaging Office Carolina Méndez-Deurloo – Research Assistant Daan van der Velden – Post Processing CT Gaia Hermans – Student Assistant Ilva van Houwelingen – Process coordinator Imaging Office Isabelle Klapwijk – Student Assistant Ivar Jole – Research Assistant Jessica Wijngaarden – Data Manager Jon Bregman – Process coordinator Imaging Office Kaoutar Ballachi – Student Assistant Laurens Groenendijk – Data manager, Research Assistant Leontien Heiligers – Coordinator Imaging Trial Office Melissa Doha – Student Assistent Miranda Slotboom – Trial Monitor Mohamed Sheikh - Student Assistant Nicole Vos van Avezathe – Research Assistant Renée Broeren – Foekens – Research Assistant Sharida Ibrahim – Administrative Assistant Victoria Ehret – Data Manager

scientific report 2025 | RESEARCH STAFF

Departmental Operational Staff Including partially appointed to research Anita Harteveld – Technical Physician Christian Cutitei – Student Assistant Dennis Kuijper – Coordinator Research & Innovation MNW Jean-Baptiste van Aarsen – Coordinator Research & Innovation MNAA Jeffrey Langerak – System Administrator Jip Holtzer – Staff Advisor Joël de Groen - Coordinator Research & Innovation CT Luud Rijnen – Coordinator Research & Innovatie MRI Mart Rentmeester – System Administrator Maureen van Duin – Staff Advisor Michelle de Bloeme - Hus – Coordinator Research & Innovation Interventional Piotr Wielopolski – MR Physicist Sylvia Bruininks – Coordinator Research & Innovation MRI Thom Korthals – Student Assistant Yoelle Kilsdonk – Staff Advisor

Erasmus MC Support Asli Mumcu – HR Officer Daphne Jerphanion – Legal Counsel Fenna de Kruif – Financial Administrator Fridjof Berdowski – Financial Advisor Karin Ter Meulen – Boer – Themacontroller Marjolein van Laere – Legal Counsel Natasja Gouweleeuw – Advisor Finance Selma de Vries – Advisor HRM Sonja Anker – HR officer Tim Malherbe – Advisor HRM

External Support Staff Mika Vogel – MRI Scientist GE Healthcare

RESEARCH SUPPORT

The Department of Radiology & Nuclear Medicine consists of two main sections: Patient Care and Research & Training. Research & Training is led by Monique de Waard, Director of Research & Training. She oversees the strategic, financial, and organizational aspects of research and is the main contact for research-related advice and legal matters. She also supports project management and prepares management reports. Lieke Visser supports Monique and helps researchers with organizational questions. Staff advisors Jip Holtzer, Maureen van Duin, and Yoëlle Kilsdonk provide project support when needed. In addition, Fenna de Kruif, Fridjof Berdowski, Natasja Gouweleeuw, and Tim Malherbe support the department with project management, financial administration, and human resources. Together, the research support team helps researchers with organizational, financial, legal, ethical, and administrative matters, allowing them to focus on their research.

 18 New PhD Students 131 Total PhD Students 15 Dissertations

Studies



21 Clinical studies started 84 Clinical studies 80 Service projects started 353 Service projects



568 Scientific Publications

External Subsidies

€ € 2.7M Government € 1.3M Charity € 2.5M Industry Monique de Waard

The Research Committee forms the center of all research activities of the department and meets once every two months. Members of the committee are full professors, associate professors and assistant professors and are leading a research group as Principal Investigator. In 2025, 30 research groups were organized across four research focus areas (Figure 1).

Figure 1: The individual research lines (30) are organised within four main research focus areas.

Research groups are led by tenured principal investigators, have their own strategic plans, substantial external funding, and at least two PhD students. The committee discusses research strategy, new opportunities, and research quality, while also promoting collaboration. Several subcommittees and working groups support topics such as research strategy, data management, scientific integrity, and communication. The committee gets advice from several working groups, who, for example, prepare policy documents, communication items or analyze output factors.

PhD students are formally appointed within Research & Training and operationally embedded within research groups. Their progress is regularly reviewed by a Research Committee subcommittee, focusing on research progress, education, thesis planning, research integrity, and data management. To strengthen connections between PhD students and the committee, an annual PhD Student Dinner is organized. In 2025 this dinner was held at the restaurant ‘BlueCity’ in Rotterdam.

Research support is provided by several groups:

Imaging Trial Office (ITO) The ITO supports clinical research by preparing ethics applications, coordinating with review boards, managing study logistics, collecting and processing data, performing image analysis, advising on regulations, and monitoring study quality. The team also includes specialists in data management and clinical trial monitoring. The data manager specializes in data safety and privacy, and development of (clinical trial) databases, which extends the level and range of support offered. The clinical trial monitor oversees the conduct of clinical trials and ensures that these trials are conducted according to protocol, Good Clinical Practice, Standard Operating Procedures and regulatory requirements.

Research technicians Research technicians support preclinical research and animal studies and perform histological, radiochemical, molecular, and imaging procedures.

Research Radiographer Research radiographers and nuclear medicine technologists collect imaging data, support new technologies, scan study participants, and perform image post-processing for research and clinical projects.

Coordinators Research & Innovation Each imaging modality has a Coordinator Research & Innovation responsible for organizing research support and translating research findings into clinical practice. In 2025, six coordinators supported MRI, CT, Intervention, and Nuclear Medicine.

ICT administrators ICT staff maintain the department’s PACS infrastructure and a wide range of clinical and research software, ensuring reliable support for large-scale imaging studies.

Biomedical engineers Biomedical engineers support the acquisition, installation, testing, and validation of imaging equipment, ensuring image quality, safety, and reliable research data.

ERGO and Generation R The medical student team of Generation R and ERGO supports our research organization. The team supports large population-based imaging studies. They perform MRI scans, assist with data collection, and conduct additional assessments such as movement tests, walking tests, and neurological screenings.

scientific report 2025 | RESEARCH SUPPORT

Research strategy and & Targets

The focus of Erasmus MC’s research in the coming years will be on socially driven research. Four goals have been formulated, all of which are in line with Koers28 and the core values of Erasmus MC: connecting, responsible and entrepreneurial.

Strategic research goal 1:

Target 4. Development of novel personalized targeted therapies for oncological diseases by studying cellular and molecular targets for diagnosis of disease and radionuclide therapy in a non-invasive manner.

Strategic research goal 2:

The Erasmus MC will develop innovative strategies to promote health by preventing disease, disease progression, and the consequences of disease. The Erasmus MC will unravel the mechanisms that are associated with a healthy life course and involved in disease, and applies this knowledge in new interventions.

Target 5. Develop and validate tools and platforms for integration of radiological imaging, pathologic imaging and laboratory exams utilizing the multi-modal data to predict outcomes and treatment responses for personalized medicine.

Strategic research goal 3:

The Erasmus MC will take the lead in the development of strategies for dealing with emerging health threats.

Strategic research goal 4:

The Erasmus MC will develop innovative methods and technologies that contribute to tailored healthcare, inclusive accessibility and sustainable healthcare.

Based upon Erasmus MC’s research strategy the department defined seven research targets for the coming years.

Target 6. Develop robust imaging biomarkers to elucidate disease etiology and to identify targets for early (lifestyle) interventions for disease prevention.

1 Target 1. Development and validation of new 7 imaging techniques that result in rapid and accurate diagnostics which are cost-effective Target 7. Development, implementation and improvement of (minimally invasive) imageand sustainable. guided treatments in clinical pathways to treat or reduce the effects of disease.

Target 2. Development and validation of new quantitative imaging biomarkers which will be used for grading disease, monitoring disease progression, and assessment of the effects of treatment.

3 Target 3. Development, validation and implementation of artificial intelligence for capacity planning, image acquisition, automated image analysis and interpretation, and creation of image reports aiming at increasing productivity and reducing costs. 25

Imaging Office

Infrastructure Team

In the rapidly evolving field of medical imaging, access to advanced technologies and expertise is essential. The department’s Imaging Office provides services ranging from data storage and image analysis software to clinical implementation, serving as a gateway to the department’s combined knowledge and skills.

Large-scale medical studies present technical and administrative challenges. The BIGR Infra team develops IT infrastructure to address these challenges and make medical imaging research more reproducible, robust, and consistent. Our infrastructure supports local projects at Erasmus MC, national initiatives, and international collaborations. The team also hosts the XNAT medical imaging archive for Erasmus MC and Health-RI.

The Imaging Office also aims to reduce the workload of imaging professionals, such as radiologists, by identifying needs and developing solutions, including automated image analysis and technician-performed radiological measurements (e.g., liver volume, aorta diameter).

We provide software and infrastructure that support researchers and collaborate with partners within and beyond the department to develop effective solutions. Using a modular approach, we have created a reference IT infrastructure that can be configured to meet the needs of different medical imaging studies and data analysis workflows. The figure provides a schematic overview of this infrastructure.

To bridge the gap between research and clinical practice, the office integrates regulatory requirements, such as the MDR and AI Act, early in the development of imaging software.

Projects The Imaging Office service portfolio can roughly be divided into Digital Image Processing, Data services (e.g. storage), and Software Development. In 2025, the Imaging Office actively worked on 19 projects, with requests originating from a variety of locations.

2025 Highlights To be able to showcase the work being performed at the Imaging Office outside of the Erasmus MC, a public website was created, see QR code. The website showcases the full-service portfolio, that also provides some examples of projects that have been worked on in 2025.

Figure: Reference infrastructure for handling data and analysis in projects and studies involving medical imaging data.

In collaboration with the ICAI stroke lab, a symposium was organized on the topic: “What should researchers who develop software know about the MDR (Medical Device Regulations)”. The symposium took place on the 4th of June 2024 and triggered a lot of interest throughout the Erasmus MC. It initiated collaborations and working groups on the topic of MDR.

scientific report 2025 | RESEARCH SUPPORT

Notable achievements –

Supported 18 projects through image analysis pipelines, study design advice, data management planning, and Imaging Office data services.

–

Developed HealthDCAT-NL with Health-RI, a metadata model that improves the findability of imaging data. [EuCanImage, EUCAIM, Health-RI]

–

Created a tool to populate FAIR Data Points from data stored in XNAT. [EUCAIM, Health-RI]

–

Developed an XNAT–Galaxy integration to enable pathology and genetics analysis of XNAT data. [PATH2XNAT]

–

Collaborating with Health-RI to prepare the Netherlands for the European Health Data Space.

–

Awarded an Open Science NL grant to develop a FAIRbased Medical Imaging Curation and Annotation Platform, improving standardization, interoperability, and accessibility.

–

Contributing to the NWO Large-Scale Research Infrastructure National Roadmap 2026–2031.

–

Developing an MDR-certified lung disease quantification application with LungAnalysis, featuring workflow management and partly machine-learning-driven annotation.

Infrastructure Team Adriaan Versteeg Cristian Cutitei Ilva van Houwelingen Marcel Koek Aisha Goedhart Hakim Achterberg Jessica Wijngaarden Viktoria Ehret Alexander Harms Henri Vrooman Jon Bregman Arsam Norouzi Ivan Bocharov Mahlet Birhanu

Population Imaging Flagship Node Rotterdam Euro-BioImagingThe European imaging community EuroBioImaging provides open access to imaging technologies, training, and data services through 40 Nodes and 330 facilities across 18 countries. One of these is the Population Imaging Flagship Node Rotterdam, embedded within the Department of Radiology & Nuclear Medicine at Erasmus MC. The Node leads developments in medical imaging research infrastructure, addressing challenges in collecting, anonymizing, structuring, storing, sharing, annotating, processing, analyzing, and integrating imaging data. Through the Node, the Imaging Office can provide services at a transnational level. The Node also plays a leading role in European infrastructure projects for medical imaging research, including EuCanShare, EuCanImage, EOSC4Cancer, and EUCAIM.

IMAGING FACILITIES

Magnetic Resonance Imaging Brand

Equipment

Year of acquisition

Location

GE Healthcare

3.0T Signa Premier

2012

Sophia

3.0T Signa Premier

2023

Central Hospital

3.0T Signa Premier

2018

Central Hospital

1.5T Signa Explorer

2016

Sophia

1.5T Signa Artist

2023

Central Hospital

1.5T Discovery MR450W

2011

Central Hospital

1.5T Signa Artist

2018

Central Hospital

1.5T Signa Artist

2018

Central Hospital

1.5T Signa Explorer

2019

Population Imaging Center

X-Ray Computed Tomography Brand

Equipment

Year of acquisition

Location

Siemens

Naeotom Alpha Photon Counting CT

2024

Sophia

Somatom Definition Edge Twinbeam

2016

Central Hospital

Somatom Definition DRIVE

2016

Central Hospital

Somatom Definition Edge

2018

Central Hospital

Somatom Definition Edge Plus

2017

Central Hospital

Naeotom Alpha Photon Counting CT

2021

Central Hospital

Somatom On.Site

2022

Central Hospital

Angiography, Interventional Radiology, and Fluoroscopic Imaging Brand

Equipment

Philips

Allura Xper FD 20/10

2010

Central Hospital

Siemens

Axiom Artis Zee MP

2018

Central Hospital

Axiom Artis Zee MP

2009

Sophia

Luminos Lotus Max

2023

Central Hospital

Artis Q-Ceiling

2018

Central Hospital

Artis Icono Biplane

2021

Central Hospital

Adora DRFi

2024

Central Hospital

Canon

Year of acquisition

Location

Single Photon Emission Computed Tomography (SPECT)-based Imaging Brand

Equipment

Siemens

Symbia T16 SPECT-CT

Year of acquisition 2011

Central Hospital

Starguide

2023

Central Hospital

Location

Mammography Brand Hologic

Equipment

Year of acquisition

Location

3Dimensions

2017

Central Hospital

3Dimensions

2018

Central Hospital

Affirm Prone Biopsy system

2017

Central Hospital

Ultrasonic Imaging Brand

Equipment

Number

Year of acquisition

Philips

Epiq 5

2014

Central Hospital

Epiq Elite

2022

Central Hospital

Epiq Elite

2022

Sophia

Epiq 7

2019

Central Hospital

My Lab Twice

2015

Central Hospital

Esaote

Location

Support Equipment Brand

Equipment

Number

Year of acquisition

Location

Medrad

Injector MR Xperion

2016

Sophia

Injector MR Xperion

2018

Central Hospital

Injector Solaris (EP)

2013

Central Hospital

Injector Mark 7 Arterion

2011-2014-2017

Central Hospital

Injector Stellant Dual

2011

Central Hospital

Injector CT Motion

2023

Central Hospital

Injector CT Motion

2023

Sophia

Injector CT Motion

2024

Central Hospital

HS Medical

Amica RF System

2023

Central Hospital

TEMA sinergie RAD inject

2019-2024

Central Hospital

Photo Camera Equipment Brand

Equipment

Number

Year of acquisistion

Nikon

D750 photo camera

2015

Location All

Z8 photo camera

2025

All

DI3D

3D equipment face

2018

Central Hospital

3dMD

3D equipment head

2019

Sophia

Einscan

Handheld 3D scanner

2020

Central Hospital

Broncolor

Lightning system - Siros 400s

2025

All

Broncolor

Lightning system - Siros 800s

2025

All

Broncolor

Lightning system - Minipuls D160

2010

Central Hospital

Quest Medical Imaging

Quest Spectrum platform 2.0

2020

Burn Center Beverwijk/ AMIE facility

Positron-Emission Tomography (PET)-based Imaging Brand

Equipment

3T SIGNA PET-MR

Year of acquisition

OMNI Legend 32

2025

Central Hospital

Siemens

Biograph mCT 128 PET-CT

2011

Central Hospital

2019

Location Central Hospital

DEXA systems Brand

Equipment

iDEXA Dual-Energy X-ray Absorptiometry System

Year of acquisition

2014

Location Central Hospital

Information & Communication Technology Brand

Equipment

Location

Associated modality

Medis

Medis Suite MR

2016

Central Hospital

MRI

Philips

IntelliSpace Portal

2015

All

CT, MRI

Scintomics

Labeling software

2014

Central Hospital

Robotica Robot Synthesizer

Labeling software

2015

Radiochemistry

Robotica Robot Synthesizer

GE Healthcare

AW Server

2012

All

MRI

Siemens

SyngoVia

2012

All

SyngoVia

2011

Central Hospital

PET

TEMA Sinergie

Year of acquisition

Dispensing software

2011

Central Hospital

Dispensing robot and Dose calibrators

Dispensing software

2024

Central Hospital

Dispensing robot and Dose calibrators

Application Server

2011

Central Hospital

SPECT

Application Server

2022

Central Hospital

SPECT

Gold3 PACS

2011

Central Hospital

SPECT

Gold3 PACS

2011

Central Hospital

SPECT

Comecer

IBC Hotlab Management Software

2008

Central Hospital

Dose calibrators

Merge

CADSTREAM

2006

Central Hospital

MRI

Hologic

Softcopy Workstation

2016

Central Hospital

Mammography

PMOD Technologies LLC

PMOD

2017

Central Hospital

PET

Xeleris Server

2023

Central Hospital

SPECT

MIM

2025

Central Hospital

PET

Hermes

Conventional X-Ray Imaging Brand

Equipment

Siemens

Mobilett MiraMax

2016

Central Hospital

Mobilett MiraMax

2016

Central Hospital

Mobilett MiraMax

2016

Sophia

Ysio XPree

2024

Central Hospital

Ysio Max

2018

Central Hospital

Ysio Max

2018

Central Hospital

Ysio Max

2018

Central Hospital

Cios Alpha

2017

Central Hospital

Cios Alpha

2017

Central Hospital

Cios Alpha

2017

Central Hospital

Cios Alpha

2017

Central Hospital

DRX Revolution

2025

Central Hospital

DRX Revolution

2025

Central Hospital

C-arm Veradius

2011

Sophia

C-arm Pulsera

2009

Central Hospital

C-arm Unity

1998

Sophia

Digital Diagnost C90

2020

Sophia

Carestream Philips

Year of acquisition

Location

Digital Diagnost C90

2020

Sophia

Digital Diagnost C90 Flexroom

2024

Central Hospital

Triathlon Trauma DR

2014

Central Hospital

Triathlon Trauma DR

2009

Central Hospital

Triathlon Trauma DR

2009

Central Hospital

Insight FD Flex

2022

Central Hospital

Insight FD Flex

2021

Sophia

Oldelft Benelux

Planmeca ProMax 2D S3

2020

Sophia

EOS Imaging

EOS

2020

Sophia

Oldelft Benelux

Hologic

scientific report 2025 | IMAGING FACILITIES

Laboratory Facilities Brand

Equipment

Scintomics

Robotics Module 1

Eckert & Ziegler TEMA Sinergie

Wallac/PerkinElmer

Comecer

Number

Year of acquisistion

2015

Location Radiochemistry

Robotica Robot Synthesizer

2014

Central Hospital

Robotics Module 2

2011

Radiochemistry

Dispensing Robot

2011

Central Hospital

Actuator Dispensing Robot with Lift

2011

Central Hospital

Dispensing Robot with lift

2024

Central Hospital

Wizard 2" 2480 Automatic Gamma Counter

2015

Radiochemistry

Wizard 2” 2480 Automatic Gamma Counter

2011

Central Hospital

Wizard 2” 2480 Automatic Gamma Counter

2011

AMIE Facility

Wizard 3" 1480 Automatic Gamma Counter

<2010

Central Isotope Lab

Dose Calibrator

2004-2019

Dose Calibrator

2008

Central Isotope Lab

Dose Calibrator

2008

Radiochemistry AMIE Facility

Central Hospital

Dose Calibrator

2010

Laminar Flow Cabinet

2008-2013

Central Hospital

Isolator

2013-2025

Central Hospital

Clean air Techniek

Laminar Flow Cabinet

<2010-2011

Central Isotope Lab

Clerad

Gammaprobe

2024

Central Hospital

Metorx

Germanium Detector + Multi-channel Analyzer

2012

Radiochemistry

Interflow

BrightSpec

Germanium Detector + Multi-channel Analyzer

2011

Central Hospital

bSCAN Thin-Layer Radio-Chromotography scanner

2015

Radiochemistry

bSCAN Thin-Layer Radio-Chromotography scanne

2019

Central Hospital

HPLC-detector

2023

Radiochemistry

Alliance e2695 HPLC with a 2998 PDA detector + Canberra radioactivity detector

2015

Radiochemistry

Acquity Arc (U)HPLC with a 2998 PDA detector + Canberra radioactivity detector

2019

Radiochemistry

Acquity H-Class Ultra-Performance Liquid Chromatography (UPLC) with a 2998 PDA detector + Bpad radioactivity detector

2011

Radiochemistry

Alliance e2695 HPLC with a 2998 PDA detector + Canberra radioactivity detector

2019

Radiochemistry

Alliance e2695 HPLC with a 2998 PDA detector + Bram and Flow radioactivity detector

2019

Radiochemistry

Thermo Fisher Scientific

Liquid Chromatograph/Mass Spectrometer (LC/MS) Quantum Ultra

2017

Radiochemistry

Biotage

Microwave Biotage Initiator

2019

Central Hospital

Leica

SP8 confocal microscope

2020

Central Isotope Laboratory – faculty building

Lighthouse

airborne particle counter ApexZ

2021

Central Hospital

airborne particle counter ApexZ

2023

Central Hospital

Waters

air sampler

2024

Central Hospital

PI Medical

Handschoenintegriteit testsysteem

2022

Central Hospital

Charles River

Endosafe image reader

2018

Central Hospital

Lablogic

Radioactiviteit detector Flow-RAM

2019

Radiochemistry

Radioactiviteit detector B-RAM

2019

Radiochemistry

Focus Area

BIOMEDICAL IMAGE ACQUISITION & ANALYSIS Its focus is to develop advanced image acquisition, image reconstruction, image processing, and machine learning techniques to optimize both the acquisition and analysis of biomedical imaging data with the aim to develop novel diagnostic, prognostic, therapy planning and therapy monitoring tools, and to develop techniques to support image-guided interventions and surgery.

Juan A Hernández Tamames, PhD full professor 34

MAGNETIC RESONANCE PHYSICS IN MEDICINE Appointment in Radiology and nuclear medicine (Erasmus MC) and Imaging Physics (TU DELFT)

Top Publications 2025

Prof. Juan Hernandez-Tamames received his MSc degree in Physics from Complutense University (Madrid, Spain) in 1992. He received his PhD degree (cum laude) from Polytechnic University (Madrid, Spain) in 1999 with a dissertation about Wavelet Transforms in fMRI. Between 1999 and 2002 he obtained several academic positions as Assistant Professor at Complutense University and at Rey Juan Carlos University in Madrid. In 2000 he was visiting professor at the Institute of Psychiatry in London (King’s College of London). In 2002 he was appointed as Associate Professor at Rey Juan Carlos University. From 2004 to 2015 he was the Head of Medical Image Analysis and Biometry Lab at Rey Juan Carlos University. From 2007 to 2014 he was the head of the Electronics Department at Rey Juan Carlos University. From 2008 to 2014 he was the director of the MR Physics Group at the Queen Sofia Research Center for Alzheimer’s Disease in Madrid. From 2010 to 2015 he was faculty of the MIT program M+Vision for medical imaging training and mentoring. Since 2020 he has a double appointment in the department of Imaging Physics in TU Delft.

Nunez-Gonzalez L, E Dopper, A van der Eerden, S Abo Seada, A Boon, M Verbeek, B Bloem, F Meijer, JA Hernandez-Tamames. Quantitative MRI-based decision model for early-stage parkinsonism diagnosis: a pilot feasibility study Neuro Image: 2025; 5:1002733. Byanju R, S Klein, A Cristobal-Huerta, JA HernandezTamames, D H Poot. Myelin water imaging from accelerated 3D-GRASE acquisitions using subspace constrained reconstruction. Magnetic Resonance Materials in Physics, Biology and Medicine 2025; 5: 873-893. Lauwers I, M Capala, S Kaushik, L Ruskó, C Cozzini, JP Kleijnen, J Wyatt, H McCallum, G Verduijn, F Wiesinger, JA Hernandez-Tamames. Synthetic CT generation using Zero TE MR for head-and-neck radiotherapy. Radiotherapy and Oncology 2025; 205: 110762.

j.hernandeztamames@erasmusmc.nl

Context

agnetic Resonance (MR) physics in medicine is a rapidly evolving discipline that continuously delivers new acquisition strategies, reconstruction methods, and quantitative biomarkers capable of advancing clinical research and diagnosis. This research line aims to keep the Department of Radiology and Nuclear Medicine at Erasmus MC at the forefront of state-of-the-art MR methodology, enabling high-quality clinical research and improved patient care. The primary mission of the MR Physics group is the development, translation, and implementation of advanced MR techniques into clinical and research workflows, while supporting other research groups in the effective use of MR technology. A key objective is to improve reproducibility, robustness, and sensitivity by moving MRI beyond morphology-based assessment toward quantitative tissue characterization. Physical MR parameters, and their relationship to biological processes and pathology, provide a pathway toward truly quantitative diagnostic imaging. Our work focuses on establishing reliable quantitative biomarkers that enable pathology-specific thresholds, enhance biological interpretability, and improve the performance of radiomics and AI models.

Research Projects: Objectives & Achievements

The MR Physics group collaborates with the Hyperthermia group (Led by Dr. Sergio Curto) in the Radiotherapy department to be able to monitor the temperature during the treatment. This collaboration is embedded in a large Dutch consortium including academic center such as AUMC, Utrecht UMC, TU Eindhoven and companies such as GE Healthcare, Philips, Sensius, etc.This project has 5 PhD students, one of them, Mariana Piedade Is PhD student of Erasmus MC MR Physics group In collaboration with UMC Utrecht In MR Thermometry.

The GEM Project This project belongs to the Dutch Top Sector Life Science and Health (“Topconsortium voor Kennis en Innovatie” or TKI Life Sciences and Health). It is a consortium including GE Healthcare, TU Delft and Erasmus MC.

EU Marie Slodowska-Curie Doctoral Networks: IQ-Brain: “Quantitative Imaging in Brain” And AIPD: “Artificial Intelligence in Parkinson Disease” The MR Physics group is participating in two prestigious Marie-Slodowska Curie doctoral networks.

Figure 1. Logo of the GEM Project with the consortium partners.

The main objective is to detect Blood Brain Barrier (BBB) breakdown in brain tumors and multiple sclerosis without the necessity of gadolinium contrast agents. The PhD student in this project is Noemi Sgambelluri (page 40).

IQ-Brain, in collaboration with Dirk Poot (page 72), Stefan Klein (page 64), and AIPD in collaboration with the Neurology Department (Dr. Agnita Boom), Anke van der Eerden (Neuroradiologist), have implemented an innovative diagnosis tool using advanced MRI biomarkers to provide earlier diagnoses of parkinsonisms. The next figure shows the criteria to differentiate among Parkinson, parkinsonisms and healthy subjects.

CARES Project: Development of personalized MR-guided thermo-chemotherapy for breast conserving surgery The integration of Magnetic Resonance (MR) imaging in the development of personalized thermo-chemotherapy for breast conserving surgery using liposomes is a transformative advancement in breast cancer treatment. MR imaging plays a pivotal role in this innovative approach, offering unparalleled precision in targeting and monitoring the treatment directly at the tumor site. By leveraging the detailed imaging capabilities of MR, clinicians can visualize the tumor in real-time, ensuring that the liposome-delivered chemotherapy and heat treatments are administered with utmost accuracy.

Figure 3. Biomarkers involved in differentiating parkinsonisms from Parkinson and Healthy subjects at early stages.

Figure 4. Pons-to-midbrain ratio to differentiate Parkinson from parkinsonisms. This figure shows the automatic segmentation of relevant anatomical features to distinguish between Parkinson and Parkinsonisms.

Figure 2. Summary of the CARES project with the consortium partners.

scientific report 2025 | MAGNETIC RESONANCE PHYSICS IN MEDICINE

ICAI LAB: Smart, adaptive MR protocols for precision diagnosis The MR physics group is actively participating in two of the five workpackages of the ICAI-LAB “MRI trustworthy AI” in collaboration with GE Healthcare. We contribute in 2 PhD theses: Alireza Samdifardheris (work package 1, page 75) and Shishuai Wang (work package 2, page 75). In this project, In collaboration with GE Healthcare (Mika Vogel and Florian Wiesinger) we are pushing the acceleration to 30 seconds multiparametric mapping in high spatial resolution to enable ultrafast clinical protocols.

quantitative MRI techniques that characterize blood– brain barrier integrity, tissue microstructure, and inflammation through intrinsic physical parameters. Building on advances in relaxometry, QSM, diffusion, and hybrid-state acquisitions, I aim to establish fast, high-resolution multiparametric protocols that can be integrated into routine clinical workflows and multicentre studies. Artificial intelligence will play a complementary role, enabling synthesis of clinically relevant contrasts, improved robustness across scanners, and extraction of biologically meaningful patterns from quantitative maps. A key priority is ensuring reproducibility, harmonization, and standardization across vendors and institutions to facilitate translation into clinical trials and precision medicine. I expect this work to contribute to earlier diagnosis, improved disease stratification, and therapy monitoring in conditions such as multiple sclerosis, neurodegeneration, brain tumours, and Parkinsonian disorders. Ultimately, the goal is to transform MRI into a predictive, quantitative decision-support tool that strengthens radiological interpretation and supports more personalized, mechanism-based patient management.

Funding Curto Ramos, Sergio, Marteen Paulides, Clemens Bos, and Juan Hernández Tamames, et al. NWO Perspective: ‘Development of personalized MR-guided thermo-chemotherapy for breast conserving surgery (CARES) Conserving the breast by heating the tumour’. 2024-2028

Figure 5. AI in acquisition and reconstruction (ICAI-LAB).

Lung MRI Project In collaboration with Dr. Pier Luigi Ciet (Radiologist, page 196), Piotr Wielopolski, PhD (Medical Physics) and the PhD student Cristina Cretu (page 201, Pier Luigi Section), we are implementing and optimized protocol for quantitative Lung MRI.

Hernández Tamames, Juan, Dirk Poot, Marion Smits, and Frans Vos TKI Match 2024: ‘GEM Project: Gadolinium Free Enhancement in MRI’. 2024-2028 Poot, Dirk, Stefan Klein, and Juan Hernández Tamames EU Marie Slodowska-Curie Doctoral Networks: ‘IQ-Brain’. 2024-2027 Van der Lugt, Aad, Dirk Poot, Stefan Klein, Marion Smits, Edwin Oei, and Juan Hernández Tamames ‘ICAI LAB Robust Phase I’. 2024-2027 Hernández Tamames, Juan, Anke van Eerden, Elise Dopper, and Agnita Boom EU Marie Slodowska-Curie Doctoral Networks: ‘APID’. 2024-2027

Figure 6. Lung T1 mapping in free breathing.

Invited Lectures

Expectations & Directions To consolidate MRI as a fully quantitative, biology-driven imaging modality capable of delivering clinically actionable biomarkers across neurological disorders. A central objective is to reduce or eliminate the need for gadoliniumbased contrast agents by developing multiparametric quantitative MRI techniques that characterize gadolinium -based contrast agents by developing multiparametric

Juan Hernandez Tamames. ‘Scan Less, See More: AI Strategies for Accelerating MRI’. GE Healthcare Symposium, ESMRMB, Marseille, France. Sept 2025.

Additional Personnel Mika Vogel - Principal Scientist GE HealthCare

Assistant Professor

Gyula Kotek, PhD Imaging physics and technology, PET/MRI center of competence Email g.kotek@erasmusmc.nl

Gyula Kotek received his MSc in Physics at Eötvös Loránd University, his PhD in Physics at University of Szeged. He did post-graduate work at various research institutes and universities in Hungary, USA and Germany. From 2003 he works in multimodal medical research and medical physics. He joined Erasmus MC in 2008. His expertise is in MRI and PET/ MR Imaging Physics from theoretical work to engineering and applications in the clinical environment.

Leadership and mission Gyula directs the Hybrid Imaging and Theranostics (HIT) Center of Competence, formerly known as the PET/ MRI CoC. Working closely with clinicians and clinical researchers, the unit has established a robust research portfolio spanning technical innovation and clinical application. Research Focus and Achievements The HIT CoC pursues cutting-edge technical initiatives including motion correction and anatomy-guided PET reconstruction, synthetic patient cohorts, advanced image reconstruction algorithms, and dynamic multimodal processing. These capabilities support diverse clinical research lines-musculoskeletal pain imaging, head-and-neck oncology, pulmonary fibrosis, and hepatocellular carcinoma-with a strategic emphasis on developing techniques that translate across disease domains. In 2025 alone, the unit mentored over ten MSc students from Delft University of Technology, VU Amsterdam, and Utrecht University through internships and thesis projects, strengthening the next generation of imaging researchers. Strategic priorities Looking ahead, the HIT CoC faces two primary challenges: consolidating current research initiatives through dedicated funding and expanding technical expertise to total-body PET/CT imaging. Realizing these goals requires sustained investment in team development and infrastructure updates. The unit remains committed to fostering collaborative partnerships internally and externally, continuing to support MSc students, and-critically-securing research funding to establish PhD opportunities alongside clinical and technical collaborators.

Figure 1. The organizational structure of the CoC ensures that it is smoothly integrated into clinical research and clinical practice. It adds focus on technical aspects required in the diverse multi-modal field. The CoC channels demands and requirements from clinical research to technical staff, and technical capabilities to clinical research and practice. The core team from left to right: Anita Harteveld, Marcel Segbers, Esther Warnert, Gyula Kotek, Rianne van der Heijden and Jukka Hirvasniemi

scientific report 2025 | MAGNETIC RESONANCE PHYSICS IN MEDICINE

Post-doc

Kars van der Weijden, PhD Advance Neuroimaging Techniques

Dr. C.W.J. (Kars) van der Weijden focuses on the development and application of advanced neuroimaging techniques to characterize microstructural brain pathology, with particular emphasis on myelin imaging, neuroinflammation, and quantitative MRI and PET biomarkers in neurological disease. His work concentrates largely on multiple sclerosis (MS), where he investigates imaging strategies capable of quantifying myelin content and lesion heterogeneity using multimodal approaches combining MRI, diffusion modelling, and PET tracers such as [¹¹C]MeDAS. Additional research lines include the study of radiotherapy-induced brain changes, brain network alterations, and the application of advanced statistical and machine learning methods (e.g., PCA-based pattern analysis) to differentiate disease phenotypes and improve biomarker sensitivity. Overall, his research aims to improve the biological specificity, reproducibility, and translational value of neuroimaging biomarkers for diagnosis, disease monitoring, and treatment evaluation in neurological disorders.

PhD Students Krishnapriya Venugopal, MSc

Xiaoyu Zhang, MSc Advisors Juan Hernandez Tamames, Florian Wiesinger, Anke van der Eerden & Elise Dopper

Advisors Juan Hernandez Tamames, Matthias van Osch, Dirk Poot & Esther Warnert

Early-Stage Diagnosis of Parkinson’s Disease and Atypical Parkinsonism using Quantitative MRI Biomarkers and Artificial Intelligence

Multi-echo-based Hybrid-EPI (HEPI) technique for measuring R2’ Measuring magnetic susceptibility changes in the brain using MRI is invaluable in the study of normal brain physiology and tumors. R2’ is an important susceptibility measurement, sensitive to the deoxyhaemoglobin of brain, enabling information about oxygenation. We propose a new R2’ measurement method using multi-echo based HEPI (that combines both GRE and SE) and study its sensitivity to changes in brain oxygenation using a respiratory challenge MRI technique.

Early differentiation of Parkinson’s disease and atypical parkinsonism remains challenging, especially in clinically unclassifiable cases (CUP). This project aims to leverage quantitative MRI biomarkers and artificial intelligence to develop a non-invasive, accurate, and clinically applicable framework for early diagnosis and improved classification of parkinsonian disorders.

Ilaria Neri, MSc Noemi Sgambelluri, MSc

Advisors Juan Hernandez Tamames, Paola Scifo, Dirk Poot & Laura Nuñez Gonzalez

Advisors Juan Hernandez Tamames, Marion Smits, Frans Vos & Dirk Poot

A multicenter investigation to assess the reproducibility and repeatability of quantitative MRI multiparametric maps

Advanced MRI and AI algorithms for GadoliniumFree Enhancement

Multiparametric (MP) quantitative MRI (qMRI) provides simultaneous acquisition of multiple qMR maps in a shorter scanning time compared to the common qMR techniques. In collaboration with the IRCCS San Raffaele Hospital (Milan, IT), we are evaluating the myelin water fraction (MWF) maps in patients affected by leukoencephalopathy obtained from the 3D GRASE and MP-qMR sequences (such as MRF, QTI and MAGIC). Our aim is to optimize and evaluate these sequences in terms of reproducibility and repeatability in different study centers.

For more than 30 years, invasive intravenous gadolinium-based contrast agents have been used in magnetic resonance imaging (MRI) to detect brain lesions. This project aims at eliminating the need for gadolinium, exploiting recent years advancements in MRI technology and Artificial Intelligence.

Mariana Piedade, MSc Advisors Clemens Bos, Juan Antonio Hernandez-Tamames, Wilbert Bartels, Roel Deckers

Development of MR-thermometry dedicated to thermotherapy of the breast MR thermometry is routinely applied to monitor tumour ablations. However, using it for thermotherapy of the breast comes with specific challenges. This project aims to tackle those challenges, in a collaboration between UMC Utrecht and Erasmus MC.

scientific report 2025 | MAGNETIC RESONANCE PHYSICS IN MEDICINE

Marcel van Straten, PhD assistant professor 42

PHYSICS IN CT TECHNOLOGY Top Publications 2025

Marcel van Straten (1974) studied Applied Physics at Delft University of Technology. His MSc project was on contrast agents in MRI. He investigated the relationship between magnetization relaxation times and concentration of contrast agents. His PhD project at the Academic Medical Center in Amsterdam focused on the application of image registration techniques in spiral CT. One of the applications provided a fully automatic technique to remove obscuring bone structures and calcifications from CT angiography images. After that he was a postdoctoral researcher at the Institute of Medical Physics of the University of Erlangen-Nuremberg. He worked on the optimization and dosimetric aspects of CT. In 2008 he joined the Department of Radiology of Erasmus MC. In 2013 he has been granted certification as a medical physics expert.

Van der Bie J, T van der Laan, M van Straten, R Booij, D Bos, ML Dijkshoorn, A Hirsch, EHG Oei, RPJ Budde. Photon-counting CT: An updated review of clinical results. European Journal of Radiology 2025; 190:112189. Van der Bie J, BP Berghout, RPJ Budde, J Gutierrez, M van Straten, D Bos. Intracranial Calcification Detection; a Comparison Between Micro-CT, Conventional CT and Ultra-high-resolution PhotonCounting Detector CT. Academic Radiology 2025; 32:3599-3607. Huijskens S, M van Straten, PC Lopes, L Rossi, P Wohlfahrt, M Hoogeman. Photon-counting computed tomography for stopping power ratio prediction in proton therapy. Medical Physics 2025; 52:e70146.

marcel.vanstraten@erasmusmc.nl

Context

n CT, image quality is influenced by many factors. It starts with the scanner’s hardware and acquisition protocol and ends with the reconstruction technique and post-processing techniques applied. In clinical practice, the radiologist would like to have the best possible image quality at the lowest radiation dose possible for a specific diagnostic task.

Besides the need for optimization of the acquisition and reconstruction technique, there is a need for a smooth introduction of these optimized techniques into clinical practice. Our research focuses on the standardization and optimization of image quality in x-ray computed tomography (CT) based on the laws of physics and driven by technological innovations.

Research Projects: Objectives & Achievements

Expectations & Directions

Evaluation of technological innovations New acquisition hardware or new reconstruction methods always claim to improve the image quality of CT scans or to reduce the dose without affecting the image quality. In order to assess the value of newly introduced acquisition and reconstruction techniques, it is of utmost importance to use objective evaluation methods to bring forth both benefits and limitations of these developments. Our research focuses on the development and application of such evaluation methods. In 2025, research continued on the dual-source photon-counting detector-based CT scanner (NAEOTOM Alpha, Siemens Healthineers). PhD-student Bas Meyer Viol startedworking in the field of cardiovascular photon-counting CT. PhD-student Thom van der Laan continued working in the field of both musculoskeletal and cardiovascular photon-counting CT.

Building a knowledge base After a successful evaluation and optimization of the acquisition and reconstruction technique, there is the need for a smooth introduction into clinical practice. This can be a complex optimization process. Changing systemproperties, acquisition parameters, or reconstruction parameters, will influence both radiation dose and imagequality. Optimization via a human interface might be time consuming and error prone. In 2025, the RHYTHM project started. It aims to improve the quality and radiation safety of medical imaging for children, adolescents, and young adults. Our research in RHYTHM focuses on guidelines for dose and image quality optimisation in diagnostic CT, and development of acquisition protocols. Moreover, Erasmus MC contributes to the development of a European dose and image repository as a tool for optimisation of paediatric diagnostic CT exams. RHYTHM is a collaboration between 20 partners from 12 European countries. More information on the countries and partners involved can be found on the official RHYTHM web page.

We use the knowledge generated in our research group to build a knowledge base on the performance of a CT scanner in various situations that we use for the automation of the CT operating procedure. Our goal is that the automated procedure makes the best use of the latest technological innovations for a given diagnostic question.

In CT, objective quantification of the performance of new technology and algorithms will allow us to determine their impact on the image quality and thus on the diagnostic performance. The knowledge obtained by the studies described above will allow for building a knowledge base that can be used for the development of a ‘knowledgeable CT scanner’ and for the automation of the CT operating procedure.

Invited lectures Marcel van Straten. ‘Protocol optimization, radiation risk in perspective and communication’. European Course of Pediatric Radiology, Rotterdam, the Netherlands. Nov 2025.

Highlights Judith van der Bie successfully exhibited her study on isotropic high-resolution photon-counting detector CT angiography. She demonstrated the feasibility of a robust approach for simultaneous assessment of inplane and outof-plane spatial resolution utilizing a spherical phantom (Figure 1).

Figure 1. A polyoxymethylene sphere was mounted on an attachment piece, which was tightened in a screw thread of the cardiac motion phantom. To avoid the presence of the attachment piece in the analyzed plane, the attachment piece was tilted by 45 degrees.

Additional Persnonnel Marcel Dijkshoorn - Research Technologist CT Sarah Stam - MSc student TU Delft

scientific report 2025 | PHYSICS IN CT TECHNOLOGY

Post-doc

Ronald Booij, PhD Bone microarchitecture assessement using photon-counting detector CT

Accurate measurements of trabecular bone microarchitecture are required for a proper assessment of bone fragility. Photon-counting detector CT (PCD-CT) has different technical properties than conventional CT, resulting in higher resolution and thereby potentially enabling in-vivo measurement of trabecular microarchitecture. High-resolution peripheral quantitative CT (HRpQCT) is considered the standard technique for in-vivo assessment of bone microarchitecture; yet, its clinical application is currently restricted by the limited availability of the scanner. PCD-CT is expected to become more available during the coming years and has potential for clinical assessment of bone microarchitecture. We compared trabecular microarchitecture measurements of cadaveric distal radii and tibiae from PCD-CT at various radiation doses with HR-pQCT and concluded that PCD-CT can accurately quantify trabecular bone microarchitecture at clinically relevant radiation doses.

PhD Students Bas Meyer Viol, MSc

Thom van der Laan, MSc

Advisors Ricardo Budde, Aart Nederveen & Marcel van Straten

Advisors Edwin Oei, Ricardo Budde, Ronald Booij & Marcel van Straten

Photon counting CT for cardio-and neurovascular imaging - Defining the new clinical standard

The Influence of PCCT on Cardiovascular and Musculoskeletal Imaging

Photon counting offers the potential for more precise evaluation of coronary artery disease, enhanced plaque characterization and clearer visualization of the in-stent lumen. PCCT may also enhance the detection and characterization of prosthetic heart valve (PHV) complications, as well as improve the identification of myocardial fibrosis through lateenhancement imaging.

Photon-Counting Computed Tomography (PCCT) en hances cardiovascular and musculoskeletal imaging by providing superior spatial resolution and tissue differ entiation at reduced radiation doses. PCCT enables im proved visualization of vascularization, stent patency, and atherosclerotic plaques, as well as detailed assessment of bone quality and implant interfaces, potentially advancing diagnostic accuracy and patient outcomes.

Marleen de Bruijne, PhD full professor 46

AI IN MEDICAL IMAGE ANALYSIS

JOINT APPOINTMENT IN UNIVERSITY OF COPENHAGEN

Top Publications 2025

Marleen de Bruijne is Professor of AI in Medical Image Analysis at Erasmus MC and University of Copenhagen. She received an MSc degree in physics (1997) and a PhD degree in medical imaging (2003), both from Utrecht University. Before joining the University of Copenhagen (2007) and Erasmus MC (2008) she was assistant/associate professor at the IT University of Copenhagen. Marleen has (co-)authored ca 250 peer-reviewed papers in international conferences and journals, holds 7 patents, is the recipient of the NWO-VENI, NWO-VIDI, NWO-VICI, and DFF-YDUN awards, and is elected fellow of the MICCAI Society. She has been program (co)chair and general co-chair of several international conferences, including MIDL, MICCAI, and IPMI. She is currently a member of the IPMI board and of the editorial board of Medical Image Analysis. Her research is in machine learning for quantitative analysis of medical images and computer aided diagnosis, with applications in pulmonary-, neuro-, and cardiovascular imaging. She has supervised 34 PhD theses in these areas.

Chen S, K Garcia-Uceda, J Su, G van Tulder, L Wolff, T van Walsum, M de Bruijne. Label refinement network from synthetic error augmentation for medical image segmentation. Medical Image Analysis 2025; 99:103355. Dudurych I, G Sidorenkov, M van Tuinen, DJ Slebos, GH de Bock, M van den Berge, M de Bruijne, R Vliegenthart. CT-based airway changes after smoking cessation in the general population. Eur J Radiol. 2025; 183:111905. Mechlenburg I, S de Raedt, HC Achterberg, M Stilling, L Rømer, K Søballe, M de Bruijne. Morphology of the dysplastic hip and the relationship with sex and acetabular version. Clinical Anatomy 2025; 38(5):512-522.

marleen.debruijne@erasmusmc.nl

Context

Research Projects: Objectives & Achievements

me and plaque composition, which are known predictors of whether a plaque is likely to rupture or not. We also investigate the value of more advanced imaging features as potential biomarkers; for instance, we found that ultrasound image texture characteristics correlate with future ischemic vascular events.

We develop novel methods for quantitative analysis of medical images. The focus is on fully automatic analyses, which makes our techniques ideally suited for large-scale imaging studies. We aim to develop techniques that are generic and that will often have multiple applications. For instance, the segmentation approach that we originally developed to segment the airway tree from CT images was later successfully applied to segment the lumen and outer wall of carotid arteries in MR and in ultrasound images and the aorta and pulmonary artery from CT; and the appearance models we developed to analyze lung texture are also used in brain structure segmentation and to derive imaging biomarkers of dementia. The different research lines are described briefly below.

Neuro image analysis

Projects: – Learning imaging biomarkers: Machine learning techniques for data-driven disease prediction. – Data-efficient and self-supervised Convolutional Neural Networks (CNNs) for biomedical image segmentation (Subhradeep Kayal)

Within neuro image analysis we develop techniques to quantify different aspects of neurodegenerative diseases, to facilitate clinical and epidemiological research in this area. We are working on techniques to automatically and robustly quantify markers of cerebral small vessel disease, including microbleeds, enlarged perivascular spaces, lacunes of presumed vascular origin, and white matter hyperintensities. We previously developed techniques to segment MR images of the brain intro gray matter, white matter and cerebrospinal fluid as well as techniques to perform brain structure segmentation as a starting point for volume and shape analysis. We have for instance shown that hippocampal shape derived from MRI scans is predictive for dementia years before clinical symptoms arise and that this provides additional predictive value over hippocampal volume.

Image analysis across imaging protocols Pulmonary image analysis Accurate and reproducible quantification of abnormalities in lung images is crucial to improve our understanding of development and progression of lung diseases, to assess the effect of treatment, and to determine prognosis in individual patients. Pulmonary image analysis at BIGR focuses on measuring structural lung damage in patients with cystic fibrosis (CF)—both in very early and in advanced stages—and on quantifying chronic obstructive pulmonary disease (COPD) from CT images. In close collaboration with researchers at the University of Copenhagen and with LungAnalysis we have developed techniques to segment and measure the dimensions of lungs, airways, and vessels, texture based methods to quantify parenchymal abnormalities, model-based image registration techniques to monitor localized changes, and motion analysis in dynamic MR and CT sequences of the breathing lungs to study respiratory insufficiency in patients with Pompe disease.

Vascular image analysis Within this research line, we develop imaging biomarkers of atherosclerosis from different imaging modalities, with the aim to improve our ability to identify patients who have a high risk of suffering a (recurring) stroke and who need surgical treatment. We worked with in- and ex-vivo MRI, CT, ultrasound and histology images, and developed automated methods to segment the carotid arteries as well as different components of atherosclerotic plaque including calcification, lipid, fibrous tissue, and intra-plaque hemorrhage. This provides measures of plaque volu-

Software to analyze medical images often stops working properly when one switches to a new scanner type or changes the imaging protocol. This limits the applicability of such software and makes it difficult to compare data from different sites. In several projects we therefore investigate how learned models can be adapted to new types of image data and how image representations can be derived that are invariant to certain changes in scan protocols, such that multi-center or multi-scanner imaging data can be analyzed more robustly.

Learning from weak labels Most machine learning approaches to quantitative image analysis need a large number of carefully, manually annotated images for model training. This requires that a) experts are not only able to assess the images visually, but also to indicate boundaries reliably, which may be problematic, for example, in case of diffuse abnormalities; and b) resources are available to perform annotation for the sole purpose of developing image analysis systems. Much more training data would be readily available if weaker labels that indicate for instance the presence, but not the location, of an abnormality could be exploited as well. In several projects we investigate how to learn from “weak” labels such as an estimate of the relative lung volume that is affected by disease as visible in CT, the number of observed lesions in a single slice in a brain MRI scan, annotated vessel diameters, or patient outcome measures.

scientific report 2025 | AI IN MEDICAL IMAGE ANALYSIS

Expectations & Directions

PhD Student

In the past few years we have turned to deep learning approaches in all our application areas. With these highly flexible models, attention to proper interpretation of which image features drive model predictions, understanding of the model’s failure modes, estimating model bias, and correction for possible confounding factors has become even more important than with more conventional learning techniques. Furthermore, in several of our studies longitudinal imaging data, where the same patient was scanned at multiple points in time, is becoming available. Although such data can be analyzed in the same way as is done for cross-sectional studies, by quantifying aspects in each individual image and then comparing the results of different time points, joint analysis of data of all time points may be more suitable to detect subtle changes.

Funding De Bruijne, Marleen with University of Copenhagen and consortium partners EU: 'Artificial Intelligence for early detection of non-communicable disease risk in people with breast cancer (ARTILLERY)'. 2023-2027 Duijts, Liesbeth, and Marleen de Bruijne Dutch Foundation for Asthma Prevention: 'Origins of childhood asthma: focus on the developmental lung structure pathway using nonradiant imaging and artificial intelligence'. 2023-2026 De Bruijne, Marleen NWO VICI: 'Learning imaging biomarkers: Machine learning techniques for data-driven disease prediction'. 2019-2028 Wendelboe, Nielsen, Olav (University of Copenhagen), Marleen de Bruijne, and consortium partners RegionH: 'Breath-CT: Diagnosing Patients Admitted with Breathlessness - Development and Validation of Machine Learning Algorithms based on Images from Computed Tomography'. 2019-2026

Additional Personnel Silas Orting - PhD, affiliated Postdoc, University of Copenhagen Shengnan Liu - PhD, associated researcher, Department of Cardiology, Erasmus MC Malte Silbernagel - PhD student, University of Copenhagen Hoel Kervadec, PhD - Postdoc, university of Amsterdam. Guest researcher at Erasmus MC

Subhradeep Kayal, MSc Advisors Marleen de Bruijne & Hoel Kervadec

Self Supervision and Data Efficiency in Biomedical Image Segmentation Convolutional Neural Networks (CNNs) have been particularly useful in medical image analysis. CNNs can be trained relatively easily to perform many kinds of image segmentation tasks, ranging from isolating abnormal tissue in a brain image to segmenting entire airway trees in a lung scan. However, learning tasks in the biomedical domain are often constrained by the lack of substantial annotated data, which is often difficult and time-consuming to obtain. In our project, we try to tackle this problem by proposing solutions on the lines of data augmentation and self-supervised learning. Recently we proposed a new self-supervision task, which is inspired by the classic blind source separation problem.

Theo van Walsum, PhD full professor 50

IMAGE GUIDANCE IN INTERVENTIONS AND THERAPY

Top Publications 2025

Theo van Walsum graduated in Informatics at the TU Delft in 1990. In 1995 he received his PhD on flow visualization at the Scientific Visualization Group of TU Delft. After one year at the LKEB (LUMC), where he picked up his interest in image processing, he became a Post-doc at the Image Sciences Institute (UMC Utrecht). There he developed his interest in improving image guidance and worked on several projects where rotational X-ray imaging was used for guidance in minimally invasive interventions. He became assistant professor at the BIGR group at the Erasmus MC in February 2005, and associate professor in 2013. He is heading the “Image Guidance in Interventions and Therapy” theme group. This group focusses on improving image guidance by integrating pre-operative image information in the interventional situation and recently started a research line in augmented reality navigation. He is also involved in the use of AI for stroke treatment and initiated the SWITCH workshop series at MICCAI. He also is one of the PIs of the Smart Surgery Lab, and Scientific Director of the ICAI Stroke Lab.

Van der Sluijs PM, R Su, BA Schoon, RA van de Graaf, ..., SAP Cornelissen. Perfusion Angiography in Reperfused Patients with Ischemic Stroke: Differentiating between Favorable and Unfavorable Outcome. American Journal of Neuroradiology 47 2026: 47: 59-65. Liu S, R Su, J Su, WH van Zwam, PJ van Doormaal, A van der Lugt. Segmentation-assisted vessel centerline extraction from cerebral CT Angiography. Medical Physics 2025; 52:e17855. Tang J, A Thabit, T van Walsum, R Marroquim, M Benmahdjoub, Manual Registration in AR-Assisted Surgical Navigation: A Comparative Evaluation. International journal of computer assisted radiology and surgery 2026; 21:147-515.

t.vanwalsum@erasmusmc.nl

Context

inimally invasive interventions are interventions where small incisions are made to diagnose or treat patients. These interventions are beneficial for patients and society: minimal trauma reduces recovery time and requires care compared to open surgery. These benefits come at the physician’s expense: direct eyesight on the anatomy of interest, such as in conventional surgical procedures, is lacking, and tactile feedback. Interventional imaging, such as fluoroscopy and ultrasound, is used to guide the physician during the intervention. These modalities come with disadvantages: ultrasound is often only available as 2D, is operator-dependent and hard to interpret; fluoroscopy only provides 2D projection images, does not have soft tissue contrast and uses ionizing radiation. They lack the wealth of information that is available from 3D diagnostic imaging, such as in CT and MRI. It is our goal to improve image guidance in these procedures, for which we follow three strategies: 1) multimodal image guidance, i.e. reliable integration of pre-operative information during interventions, 2) improved visualization in guidance, by projecting information in the physician’s field of view using modern Augmented Reality headsets, and 3) image-based support in therapeutic and interventional decision making.

Research Projects: Objectives & Achievements

Augmented reality

Multimodal image guidance Navigation approaches have become state-of-the art of brain surgery and orthopedics. Application of this technology in e.g. cardiac and abdominal interventions is hampered by continuous tissue motion and deformation. Our aim is to develop robust techniques for integrating pre-operative information that can be used in cases where tissue motion and deformation may occur. To this end, we are developing and evaluating methods to build pre-operative models for interventional planning, and for alignment of these pre-operative models with the interventional situation, by using interventional images, position tracking information and motion/deformation models. Project - Automated occlusion detection In DSA Images (Anushka Kore)

Conventional navigation approaches, where surgical instruments are shown in relation to pre-operative imaging on a 2D computer screen, result in continuous switching of focus (from screen to surgical field), and difficulties in hand-eye coordinates. Our objective is to develop and assess mixed reality approaches that allow integration of 3D imagery in the surgical field of view. Main challenges are obtaining and mainlining accurate alignment of the images with the patient. On this topic we collaborate with the departments of oral and maxillofacial surgery, neuro surgery, trauma surgery, plastic and reconstructive surgery and oncological surgery. We have developed an Augmented Reality extension of the BrainLab navigation system that allows visualization of the tools and anatomy aligned with the patient in the surgeon’s field of view. In addition, a system for incision planning for craniosynostosis, using a Hololens 2 and an electro-magnetic tracking system, has been developed and assessed in a phantom study. We are currently working towards assessment at the operating room. Projects - Augmented reality for Le Fort 1 surgery (Mohamed Benmahdjoub) - Device-agnostic framework for surgical guidance using augmented reality (Abdullah Thabit) - Manual alignment of virtual and physical models (Jiaqi Tang)

Figure 1: Occlusion detection in DSA images.

- Infrared tracking using Magic Leap (Jiaqi Tang)

Figure 2: Experimental setup for tracking retroreflective markers with infrared light using a Magic Leap 2.

scientific report 2025 | IMAGE GUIDANCE IN INTERVENTIONS AND THERAPY

Therapeutic decision making in stroke

Decision making In Surgery

In 2015, mechanical thrombectomy was proven to be an effective treatment for patients with a stroke caused by a large vessel occlusion, and thrombectomy has become the therapy of choice for stroke patients with a large vessel occlusion. Unfortunately, not all patients recover well from the stroke, even after a seemingly successful intervention. Our aim here is to use imaging (and other data) to optimally support therapeutic decision making. In the Q-Maestro project, we focus on interventional imaging, investigating whether, and to what extent, Digital Subtraction Angiography images can be used to better determine treatment effect and predict patient outcome. We are also working on approaches to integrate 3D baseline imaging data in the intervention.

Fracture detection, quantification and detection of abnormalities are important both for clinical research (e.g. to assess the added value of various surgical approaches) and for therapeutic decision making. Since a few years, there is a strong collaboration in this area with the Department of Trauma Surgery, focusing on calcaneous, distal radius and rib fractures, and with the Department of Cranio-Maxillofacial surgery, on automating 3D cephalometric analysis.

Projects - AutoTICI to clinic (Lotte Strong, Matthijs van der Sluijs, Ellen Hu)

- Automating 3D cephalometric analysis (Bram Roumen)

- Assessment of quantitative perfusion DSA with MR Perfusion (Mouad Allaoui, Frank te Nijenhuis)

Projects - 3D quantitative analysis of the calcaneous (Alexander Wakker)

- Quantitative analysis of extra-articular distal radius fractures (Hannah Winder, Evi Schotman, Zara van Raak) - Prediction of pneumonia following traumatic rib fractures (Geanne Bakker)

- Pre post thrombectormy vessel matching in DSA (Marten Leenders) - Multimodal functional outcome prediction in EVT (Frank te Nijenhuis) - Automated imaging biomarkers assessment in acute ischemic stroke (Lennard Wolff) - Optimal projection angles for DSA (Cristian Cutitei) - Artery-Vein separation in CTA (Beatriz Cardoso Mota) - Clinical decision support and EVT optimization (Xi Li)

Figure 3: Arterial change detection in DSA imaging.

Expectations & Directions

Funding

We developed several multi-modal image guidance approaches. One direction that I would like to explore is the integration of the results in the augmented reality line with the multi-modal image guidance results.

Klaver, Caroline, Theo van Walsum, and Nicolas Chateau Health Holland TKI Call: ‘O-Vision -- Adaptive Optics imaging: A guiding star to save vision’. 2022-2026

Augmented reality applications in medicine Is a quickly evolving field. In this research line, we will continue developing technology for AR guidance in interventions, such as fast and reliable registration of the image to the patient. Based on this technology, we intend to build prototype systems for using augmented reality in various surgical and minimally invasive interventions. Our stroke research is strongly linked to clinical trials within the CONTRAST consortium, and the MR Clean Registry. The ICAI Stroke Lab, part of the ROBUST program for AI in the Netherlands, has started in 2023. In this lab, we will further develop AI approaches for therapy and rehabilitation of stroke patients in a multidisciplinary setting.

Van der Lugt, Aad, Wiro Niessen, and Theo van Walsum NWO LTP ROBUST program: ‘ICAI Stroke Lab: from 112 to Rehabilitation’. 2022-2027 Molina, Carlos, ... Hester Lingsma, Bob Roozenbeek, Ruud Selles and Theo van Walsum Horizon Europa IHI: ‘UMBRELLA (Unleashing a CoMprehensive, Holistic and Patient Centric Stroke Management for a Better, Rapid, AdvancEd and PersonaLised Stroke Diagnosis, TreAtment and Outcome Prediction)’. 2024-2029

Invited Lectures Theo van Walsum. ‘AI in de zorg’. Stroke Symposium, Fransiscus Vlietland, Rotterdam, the Netherlands. May 2025.

Several quantifications for fractures have been developed in strong collaboration with the Department of Trauma Surgery. We expect to further enhance these methods, extend their application, and validate these on larger datasets.

Figure 4: Comparing planning vs results in CMF surgery.

scientific report 2025 | IMAGE GUIDANCE IN INTERVENTIONS AND THERAPY

Additional Personnel Xi Li - PhD student Adriaan Versteeg – Research Software Engineer for CONTRAST 2 project, Jan. 2025 – Dec. 2025 Ching Khan – Data Manager for CONTRAST, Jan. 2025 - Dec. 2025 Arsam Norouzi – Data Manager for CONTRAST, Nov. 2025 - Dec. 2025 Zara van Raak – Intern TU Delft, Dec. 2024 – Feb. 2025 Evi Schotman – Intern TU Delft, Sep. 2025 – Nov. 2025 Beatriz Cardoso Mota – Intern TU Delft, Sep. 2025 - Oct. 2025 Shalini Ramlagan – Intern TU Delft, Apr. 2025 - Jun. 2025 Hanieh Molkaraie – Intern University of Bologna, Aug. 2024 – Feb. 2025 Hannah Winder – Intern TU Delft, Dec.2024 - Feb. 2025 Cristian Cutitei – Comp. science MSc TU Delft, Sep. 2025 Maarten Leenders – BME MSc TU Delft, Feb. 2025 - Jan 2026 Jiaqi Tang – BME MSc TU Delft, Oct. 2024 - May 2026 Geanne Bakker – TM MSc TU Delft, Dec. 2024 - Jul. 2025 Anushka Kore – BME MSc TU Delft, Jul. 2024 -Sep. 2025 Mouad Allaoui – MSc VU, Jul. 2024 - Sep. 2025 Beatriz Louro Correia de Miranda - BME MSc University of Lisbon, Feb. 2024 – Feb 2025 Emma de Bruin – BME MSc TU Delft, Nov. 2024 - Jul. 2025 Cristo van den Berg, Mirre Blaauboer, Carlijn keppers & Daan van Erp - KT BSc TU Delft, Mrt. 2025 - Jul. 2025 Ellen Driessen - Intern TM TU Delft, Sep. 2025 - Nov. 2025 Joana Vitorino - Intern University of Lisbon, Apr. 2025 Jun. 2025

Post-doc

Mohamed Benmahdjoub, PhD Endoscopic lung tumor resection

Advancements in imaging technology enabled earlier detection of tumors. Early resection of these tumors has a significant impact on patient well-being. However, due to their small size, it can be difficult to localize them in endoscopic images resulting in partial tumor resection or over-resection, which can negatively affect lung function. Tumor localization and augmentation For breast surgery, Sirius medical® developed a magnetic probe that detects magnetic seeds. These seeds are inserted by an interventional radiologist inside or close by to the tumor to assist surgeons localize the tumor. The probe is moved by the surgeon around the seed which allows for the distance between it and the probe tip to be shown on a screen. This project has three main goals: 1- Adapt the probe and the seed to enable their use in endoscopic lung surgery. 2- Detect and estimate the 3D pose of the seed and probe. 3- Provide an augmented endoscopic image visualizing the seed location on the lung (i.e., the tumor).

PhD Students Matthijs van der Sluijs, MD, MSc

Abdullah Thabit, MSc Advisors Theo van Walsum, Wiro Niessen & Eppo Wolvius

Advisors Theo van Walsum, Aad van der Lugt & Sandra Cornelissen

PhD Obtained 9-12-2025

Quantitative Microvasculature AssEssment in projection angiography of ischemic STROke patients

Augmented Reality image-guidance in surgery Navigation has become standard of care for several surgical areas, such as neurosurgical and orthopedic procedures. However, conventional navigation systems suffer from a few drawbacks, such as the poor hand-eye coordination and the need to switch focus (from the operative field to the screen). The objective of my PhD is to investigate the use of AR with head mounted displays as an alternative to conventional systems in surgical navigation.

This project aims to develop imaging parameters that quantify perfusion restoration after reperfusion therapy using Digital Subtraction Angiography (DSA). Through this, clinical decision making of the neuro-interventionalist performing the treatment could potentially be influenced, as supplementary locoregional therapeutic action in the intervention stage might still be available.

scientific report 2025 | IMAGE GUIDANCE IN INTERVENTIONS AND THERAPY

Frank te Nijenhuis, MD, MSc

Lennard Wolff, MD, MSc

Advisors

Advisors Aad van der Lugt & Theo van Walsum

Theo van Walsum, Danny Ruijters & Sandra Cornelissen

Applications of Artificial Intelligence in Image Guided Interventions for Acute Ischemic Stroke

Automated imaging biomarkers in acute ischemic stroke

The aim of this project is to investigate how AI and computer vision techniques can be employed to enhance stroke interventions. The main focus is on improving image guidance by including pre-operative imaging information. Additionally, collaboration within the lab facilitates integration of AI methods along the entire stroke care pathway.

Several imaging biomarkers predict outcome in patients with acute ischemic stroke and effects of endovascular treatment (EVT). Automated analysis tools might decrease the variability in the evaluation of imaging biomarkers with a subsequent improvement of prediction tools.

Alexander Wakker, MSc

Bram Roumen, MSc

Advisors Michiel Verhofstad, Theo van Walsum, Jan-Jaap Visser & Mark van Vledder

Advisors Eppo Wolvius, Theo van Walsum & Bart Cornelissen

Understanding the 3D anatomy of the Calcaneus

Automation of preoperative planning and intraoperative assistence

Patients with calcaneus fractures often require complex reconstructive surgery. However, there is currently no established method to quantitative perform automated morphological measurement on 3D models of the calcaneus. Therefore, a pipeline will be developed to automatically perform morphological measurements on 3D models of the calcaneus.

By integrating digital tools to support surgical teams in preoperative preparations and intraoperative guidance, we aim to streamline OR procedures. Our goal is to reduce OR times, minimize the need for revisional surgeries, and maximize the utilization of OR resources, while improving the OR staff's preceived workload.

Jaiqi Tang, MSc Advisors Theo van Walsum & Ricardo Marroquim

AR-Assisted Surgical Navigation via IR Marker Tracking This project develops an augmented reality surgical navigation system using the Magic Leap 2 headset. By tracking infrared markers attached to surgical instruments with onboard world cameras, the system enables real-time instrument localization and visualization to support intraoperative guidance without relying on external tracking infrastructure.

Gennady Roshchupkin, PhD assistant professor 58

COMPUTATIONAL POPULATION BIOLOGY

JOINT APPOINTMENT IN EPIDEMIOLOGY

Top Publications 2025

Gennady Vasilievich Roshchupkin received his MSc from Moscow State University in Physics Department with specialization in stellar astronomy and astrophysics. After that he spent 3 years as a senior research engineer at a network security company. In 2014 he started his PhD and graduated (cum laude) in 2018 at Department of Radiology and Department of Epidemiology at Erasmus MC. He established his research line and was appointed as Assistant Professor in 2022. Gennady received several personal grants (e.g. VENI, NIH, Erasmus MC Fellowship). His research focused on developing and application of methods for the integrative analysis of large-scale biological, epidemiological and clinical data. Gennady has a broad background in statistics, computer science, machine learning, deep learning, medical image analysis and genomics. Since 2019 Gennady is chairing Machine Learning working group in (CHARGE) Consortium. Since 2021 Gennady is also leading Bioinformatics Working group in Genomics of MusculoSkeletal traits Translational Network. Gennady is co-founder of Erasmus MC squAIre (Society for Quantitative Artificial Intelligence Research). Since August 2025 Gennady works part time as Programme Manager AI Strategy for Research, Development and Innovation for UMCNL.

Deng R, E Perenthaler, A Nikoncuk, S Yousefi, K Lanko, R Schot, M Maresca, E Medico-Salsench, LE Sanderson, MJ Parker, WF van Ijcken. BRAIN-MAGNET: A functional genomics atlas for interpretation of non-coding variants. Cell 2025; 189:676-695. Liu X, B Li, MW Vernooij, EB Wolvius, GV Roshchupkin, EE Bron. AI-based association analysis for medical imaging using latent-space geometric confounder correction. Medical Image Analysis 2025; 102:103529. van Hilten A, F Melograna, B Fan, W Niessen, K Van Steen, GV Roshchupkin. Detecting genetic interactions with visible neural networks. Communications Biology 2025; 5:8:874.

g.roshchupkin@erasmusmc.nl

Context

he Computational Population Biology group is melding computational science with biological, epidemiological, and clinical data analysis to address critical health issues facing society today. Through the sophisticated analysis of largescale datasets, our team seeks to deepen the understanding of omics’ role in complex traits, aiming to revolutionize the prevention, diagnosis, and treatment of diseases. This research is not just about scientific discovery; it’s about responding to a societal need for better healthcare solutions, reducing the burden of disease worldwide, and enhancing the quality of life for individuals. Leveraging the latest in genomics, medical imaging, computer science, statistics, and machine learning, we are uniquely positioned to sift through the rich and complex data that modern healthcare generates. Our work is crucial in an era where personalized medicine and targeted therapies are becoming the cornerstone of healthcare, offering new hope and possibilities for patients.

Research Projects: Objectives & Achievements

Drug Discovery – AI and multi-omics integration facilitate the identification of therapeutic targets and optimize treatment strategies, advancing precision medicine.

Key Research Areas:

This closed-loop system ensures that population-level insights refine genetic discoveries, while breakthroughs in genetics and drug development fuel personalized interventions, creating a continuous cycle of innovation.

- Neuroimaging – AI-powered analysis of structural MRI, DTI and resting-state fMRI to map brain structure and function. - Facial Morphology – Exploring 2D and 3D facial features as phenotypic biomarkers linked to genetic and health indicators. - Musculoskeletal Health – Investigating genetic and biomechanical factors that influence bone and joint health. 3D Advancing AI-driven

Imaging in precision analysis of

Clinical Research – diagnostics through 3D medical imaging.

Projects: - Enhancing brain imaging power and genetic discovery using genetic data and high-dimensional, multimodal brain MRI measurements (Jing Yu) - Unravelling the Dark Matter of the Genome with novel AI methods (Ruizhi Deng) - Advancing knowledge of genetic diseases by unraveling complex mechanisms with interpretable machine learning techniques (Arno van Hilten)

Federated Learning & Industry Collaborations We are at the forefront of federated learning research, developing privacy-preserving AI models that enable multi-institutional collaboration without sharing raw data. Our group works with: - NVIDIA, driving advancements in AI infrastructure for healthcare; - OASYS NOW and other startups, developing innovative AI solutions for clinical and biomedical applications; - Global research consortia, integrating AI-driven insights into large-scale, multi-center studies.

Multi-Modal Data Analysis & Endophenotype Research We focus on advanced phenotyping and endophenotype analysis, integrating imaging, omics, and clinical data to unravel the etiology of complex traits.

- Multimodal imaging techniques for the identification of brain features associated with (chronic)pain and comorbidities (Xianjing Liu) - AI application for dental development (Li Chen)

Circular AI-Driven Pipeline for Precision Medicine Our research group develops and implements a circular AI-driven pipeline that integrates clinical genetics, population studies, and drug discovery to transform precision medicine. By combining advanced AI, federated learning, and multi-omics approaches, we bridge fundamental research with real-world applications, accelerating insights into disease mechanisms and therapeutic innovations.

Closed-Loop AI for Clinical Genetics, Population Studies & Drug Discovery Clinical Genetics – We leverage AI to analyze genomic data, identifying critical motifs and pathways that reveal underlying disease mechanisms.

Figure 1. BRAIN-MAGNET AI tool application for non-coding DNA mutation annotation.

Population Studies – Using federated learning, we enable secure, large-scale collaboration across institutions, harmonizing diverse datasets while preserving privacy.

Open Science & AI for Research Accessibility We are committed to open science and the development

scientific report 2025 | COMPUTATIONAL POPULATION BIOLOGY

of open-source tools that enhance data visualization, AI model accessibility, and collaborative research platforms. By fostering an open and transparent research environment, we empower scientists and clinicians to explore, validate, and apply AI-driven insights to realworld challenges. Through this AI-driven circular pipeline, we are shaping the future of precision medicine, biomedical AI, and multiinstitutional research, making healthcare innovations more accessible, efficient, and impactful.

Collaboration The group actively involved in various collaboration projects within Erasmus MC: Departments of Epidemiology Radiology and Nuclear Medicine, Neurosurgery, Plastic Surgery, Internal Medicine, Psychiatry, Neuroscience, Craniomaxillofacial surgery, Genetic Identification, Dermatology, Clinical Genetics and AI Accelerator. Also, nationally and internationally: Netherlands Cohorts Consortium (NCC), CHARGE consortium (the Cohorts for Heart and Aging Research in Genomic Epidemiology), ENIGMA consortium, eQTLGen consortium, and EADB consortium (A European DNA bank for deciphering the missing heritability of Alzheimer’s disease).

Personalized Patient Experiences’. 2021-2025 Roshchupkin, Gennady ZonMW VENI: ‘Explainable Artificial Intelligence to unravel genetic architecture of complex traits’. 2021-2025 Roshchupkin, Gennady NIH R01: ‘An integrative computational interrogation of circuit dysfunction in schizophrenia via neural timescales’. 2022-2027 Roshchupkin, Gennady, and Daniel Bos Erasmus MC, TU Delft & EUR Convergence Flagship: ‘ALIVE: A Lifecourse and Individual-based View on Lifestyle to Enhance Health’. 2022-2026 Roshchupkin, Gennady European Union’s Horizon 2020 research and innovation programme: ‘Identifying molecular mechanisms of pain related disorders’. 20222026 Roshchupkin, Gennady Erasmus MC Fellowship: ‘GenetiX: Decoding the Genetic Puzzle of Complex Traits with Federated learning and Explainable AI’. 2023-2028 Roshchupkin, Gennady Dutch Ministry of Education, Culture and Science (OCW): ‘Starting Grant’. 2023-2029

The group members contribute as AI experts in several EU COST actions networks.

Roshchupkin, Gennady, and Tareq Abdel Alim Sophia Stichtingen: ‘Machine learning for postoperative shape prediction in craniosynostosis treatment using 3D models’. 2024-2025

Expectations & Directions

Invited Lectures

Our research group is dedicated to pushing the boundaries of AI in precision medicine by:

Gennady Roshchupkin. ‘Best Young Researcher Award Lecture’. Dutch Bio-Medical Engineering Conference, Egmond aan Zee, the Netherlands. Jan 2025.

- Advancing federated learning to enable scalable, privacypreserving multi-institutional AI research. - Enhancing explainability in AI models, making genomic and biomedical AI more interpretable and actionable. - Expanding multi-omics integration, bridging genomics, imaging, and clinical data for deeper insights. - Strengthening collaborations with industry partners, startups, and research consortia to translate AI innovations into clinical applications. - Promoting AI accessibility & open science, ensuring fair and ethical AI applications in global healthcare. Omics, text, and voice analysis, among others, to create a holistic understanding of health and disease.

Funding Roshchupkin, Gennady Erasmus MC-TKI-LSH: ‘Seeing is Believing: Virtual & Augmented Reality For Enhancing

Gennady Roshchupkin. ‘Federated Learning In Healthcare’. Data & AI in Healthcare TechPros, CIC, Rotterdam, the Netherlands. Feb 2025. Gennady Roshchupkin. ‘AI-Powered Innovations: Advancing or Slowing Down Research in Oral and Maxillofacial Surgery’. OFMS Department Research Day, Rotterdam, the Netherlands. March 2025. Gennady Roshchupkin. ‘AI in Healthcare’. Pediatric Research Meeting, Rotterdam, the Netherlands. March 2025. Gennady Roshchupkin. ‘AI-Enabled Lifestyle Interventions in Cohort Studies’. Convergence Flagship Workshop, Rotterdam, the Netherlands. April 2025. Gennady Roshchupkin. ‘INSIGHT: Innovative National Synthetic Data Infrastructure for General Health and Testing’. VU, Amsterdam, the Netherlands. April 2025. Gennady Roshchupkin. ‘Integrating Multi-Omics Data With Explainable AI to Advance Understanding of Mental Health

Disorders’. Annual Meeting, Society of Biological Psychiatry, Canada. Toronto, Canada. April 2025. Gennady Roshchupkin. ‘The Art of AI-Driven Healthcare: Navigating through Modern Medicine with Ancient Wisdom’. Clinical Genetics Research Day, Rotterdam, the Netherlands. April 2025. Gennady Roshchupkin. ‘Workshop: The Use of AI in Dementia’. Alzheimer Center Workshop, Rotterdam, the Netherlands. May 2025. Gennady Roshchupkin. ‘Mission AI-possible’. Imaging on the Move Conference, Rotterdam, the Netherlands. May 2025. Gennady Roshchupkin. ‘From Models to Medicine: The Path Toward Medical Superintelligence’. AI Symposium ICAI Labs, Rotterdam, the Netherlands. July 2025. Gennady Roshchupkin. ‘Responsible AI in Research and Open Science’. ROTTERDAM Science Club, Rotterdam, the Netherlands. Oct 2025. Gennady Roshchupkin. ‘Can Artificial Intelligence Make Science More Human?’. Science Communication Community, Rotterdam, the Netherlands. Oct 2025. Gennady Roshchupkin. ‘AI Roadmap for the Women’s Health Research & Innovation Center (Erasmus MC)’. Women’s Health Research & Innovation Center, Rotterdam, the Netherlands. Oct 2025. Gennady Roshchupkin. ‘Decoding Genomics with AI’. Symposium4u, Ede, the Netherlands. Nov 2025. Gennady Roshchupkin. ‘Responsible Data Connection and Innovation: The Role of AI’. From Data to Society Conference, Rotterdam, the Netherlands. Nov 2025. Gennady Roshchupkin. ‘The Power of AI in Nuclear Medicine’. Symposium4u, Ede, the Netherlands. Dec 2025. Gennady Roshchupkin. ‘Federated and Explainable AI for Population Multi-Omics Research’. Genomics Research Infrastructure Workshop, Utrecht, the Netherlands. Dec 2025.

Additional Personnel Sara Okhuijsen – MSc student, ‘Evaluating the Privacy of Confidential Multi-Party AI (CoMPAI) on Genomic Data’. Ruizhi Deng – Visiting PhD student, Clinical Genetics Erasmus MC Yasna Afshari – MSc student, ‘Web-Based Viewer for Population-Based Studies’. Semyon Galichenko – Linux system administrator and data scientist Lennart Karssen – Senior Linux administrator

scientific report 2025 | COMPUTATIONAL POPULATION BIOLOGY

PhD Students Xianjing Liu, MSc

Arno van Hilten, MSc

Advisors Gennady Roshchupkin & Eppo Wolvius

Advisors Wiro Niessen & Gennady Roshchupkin

PhD Obtained: 14-05-2025 3D facial shape analysis Unravel the complexity involving the human face, genetic factors, environmental factors, and their connection with health outcomes, by developing and applying advanced artificial intelligence-based techniques with a special focus on addressing challenges of confounder and interpretability in deep learning models.

Interpretable Machine Learning in Genetics Interpretable Machine Learning in Genetics is a crucial approach in understanding the complex genetic mechanisms underlying various traits and diseases. It involves applying machine learning techniques to genetic data in a way that allows for clear interpretation of the models’ decisions. This can lead to insights about the significance and impact of specific genes or genetic variations. These insights can enhance our understanding of genetic diseases, enable personalized treatment strategies, and potentially facilitate the discovery of new therapeutic targets.

Zuqi Li, MSc Advisors Kristel Van Steen, Peter Claes, Nataša Pržulj & Bertram MüllerMyhsok

Li Chen, MSc

Novel multi-omics deconfounding variational autoencoders can obtain meaningful disease subtyping

Advisors Gennady Roshchupkin & Eppo Wolvius

With the availability of large-scale multi-omics data, clustering algorithms can be empowered by deep learning models to better exploit the heterogeneity between individuals. We proposed four variational autoencoder-based deconfounding approaches utilizing multi-omics data. We demonstrated our models effectively mitigated the artificially introduced confounder effect and unveiled meaningful patient subgrouping.

Dental Development in Childhood and Adolescence: Indicators, Determinants, and Novel Imaging-based Assessment Approaches This project aims to investigate dental development as a measurable developmental phenotype in children and adolescents, focusing on its role as an indicator of general growth, its association with early-life feeding and craniofacial development, its potential relationship with brainstem morphology, and the application of AI-based approaches to capture longitudinal dental developmental patterns beyond conventional staging methods.

Jing Yu, MSc Advisors Gennady Roshchupkin & Arfan Ikram High-dimensional brain MRI measurements for genetic analysis Incorporating modern techniques including deep learning, federated learning, enrichment analysis for high-dimensional multi-model brain MRI measurements and genetic data to increase the power of brain imaging and genetics analysis discovery, in addition to facilitate brain disease understanding and prediction.

Stefan Klein, PhD full professor 64

APPLIED MEDICAL IMAGE ANALYSIS

Top Publications 2025

Stefan Klein is Full Professor in Applied Medical Image Analysis and is General Chair of the Biomedical Imaging Group Rotterdam (BIGR). His interests span a wide range of domains: in the last 5 years he has worked on fundamental technology for accelerated magnetic resonance image (MRI) acquisition and quantification, multi-scale and multi-modal retina imaging, novel spatiotemporal models of the developing and aging brain, and AI-supported diagnosis and prediction methods for various types of cancer, neurodegenerative disorders, osteoarthritis, and major eye diseases. Besides performing research, Stefan is also active in setting up infrastructures that facilitate research in medical imaging, and he has for instance initiated a national Health-RI research archive for medical imaging data, currently used by numerous multi-centre imaging studies in the Netherlands, he is Imaging Community manager at Health-RI, and director of the Euro-BioImaging Population Imaging node. He strongly believes in the power of open science to promote research reproducibility: sharing code, sharing data, and collaborating rather than competing.

Spaanderman DJ, M Marzetti, X Wan, AF Scarsbrook, P Robinson, EH Oei, JJ Visser, R Hemke, K van Langevelde, DF Hanff, GJ van Leenders, C Verhoef, DJ Grunhagen, WJ Niessen, S Klein, MP Starmans. AI in radiological imaging of soft-tissue and bone tumours: a systematic review evaluating against CLAIM and FUTURE-AI guidelines. eBioMedicine 2025; 114: 105642. Bastiaansen WA, S Klein, B Hojeij, E Rubini, AH Koning, W Niessen, RP Steegers-Theunissen, M Rousian. Automatic Human Embryo Volume Measurement in First Trimester Ultrasound From the Rotterdam Periconception Cohort: Quantitative and Qualitative Evaluation of Artificial Intelligence. Journal of Medical Internet Research 2025; 27:e60887. Zhang Y, Y Zhao, H Xue, P Kellman, S Klein, Q Tao. Recurrent inference machine for medical image registration. Medical Image Analysis 2025; 106:103748.

s.klein@erasmusmc.nl

Context

he Applied Medical Image Analysis research line focuses on the development and validation of novel medical image analysis methods using advanced computational methods based on numerical mathematics, signal processing, and artificial intelligence (AI) including (deep) machine learning. We bring state-of-the-art techniques from computer science to the medical imaging domain, further developing, optimizing and rigorously validating them.

This research line is part of the Biomedical Imaging Group Rotterdam (BIGR), a collective of principal investigators at the forefront of medical image analysis & AI research, working together to develop a joint strategy for research and infrastructure and create a thriving and dynamic working atmosphere. BIGR is rooted and embedded in Department of Radiology & Nuclear Medicine, and has many collaborations with other research groups within and beyond Erasmus MC. An overview of the BIGR staff (PI’s plus support staff) is shown in Figure 1 - more information can be found at https://www.bigr.nl.

Research Projects: Objectives & Achievements Prenatal Image Analysis We have a fruitful collaboration with the Dept. of Obstetrics and Gynaecology (Dr. Rousian & Prof. SteegersTheunissen), aimed at the analysis of 3D ultrasound images of the embryo during pregnancy. In 2024, Wietske Bastiaansen defended her PhD thesis and continued as a postdoctoral researcher, focusing on modeling embryonic, fetal, and placental development using AI. Her research aims to identify adverse growth patterns and congenital anomalies. Furthermore, she explores the impact of lifestyle behaviors and various maternal and paternal factors—both modifiable and non-modifiable, such as diet, smoking, and alcohol consumption—during the periconceptional period. Marcella Zijta started her PhD project in 2023, in a collaboration with Bernadette de Bakker from the Amsterdam Medical Center, who is the founder of the Dutch Fetal Biobank. In this project, the unique data from the biobank is used in conjunction with the in-vivo ultrasound data of the Rotterdam Periconceptional Cohort, with the ultimate aim to improve the detection of congenital anomalies based on firsttrimester 3D ultrasound. PhD students Nikolai Herrmann, Anne Akerboom and Research Software Engineer Abdullah Thabit started in 2024/2025, funded by two projects

a major boost thanks to the launch of an Innovation Center for Artificial Intelligence (ICAI), in close collaboration with GE Healthcare. Dr. Dirk Poot is lab manager, Stefan Klein and Aad van der Lugt are scientific directors. In this ICAI lab, all research revolves around improving and evaluating the trustworthiness of AI methods. Four PhD students were appointed at our department: Alireza Samadi and Shishuai Wang work on the improvement of MRI acquisition and reconstruction methods (with Prof. Tamames and Dr. Poot, see p. 34 & 72); Xinyi Wan and Gonzalo Mosquera Rojas work on the development of AI methods to support diagnosis and prediction in patients with bone, soft-tissue, and brain tumors with Prof. Oei, Dr. Visser, Dr. Starmans, and Prof. Smits). A fifth PhD student, Jamie Verweij, has been appointed at Erasmus University Rotterdam, studying the trustworthiness of AI methods from a social sciences perspective. In 2025, the first results of this research have been submitted to international peerreviewed journals and presented at various conferences.

AI in Cancer Imaging The concept of personalized medicine has been fully embraced in oncology for years. Medical imaging plays a crucial role here, and many studies have shown the promise of AI to support image interpretation, disease monitoring, diagnosis and prediction. Yet a few methods have reached clinical practice. Our research in this domain includes applications to multiple cancer types and aims to accelerate the translation to clinic by tackling major barriers to adoption, in particular the lack of trustworthiness, explainability, and generalizability of AI methods. In collaboration with Prof. Smits (p. 152), we have a strong research line on AI methods for glioma (brain cancer). Four PhD students (Van der Werff, Mosquera Rojas, Dille, Van Leeuwen) and one PostDoc (Tareq Abdel Alim) are currently contributing to this line of research, all focused on development of image analysis methods for tumor segmentation, MRI quantification, vascular fingerprinting, molecular classification, tumor grading, and prognosis.

Figure 1: Biomedical Imaging Group Rotterdam (BIGR).

in collaboration with GE Healthcare: FAST-AI and AI4AI. In these projects, we focus especially on improving the robustness and explainability of AI methods for automated analysis of first-trimester 3D ultrasound images.

Trustworthy AI for MRI Our research on AI methods for the acquisition, reconstruction, processing, and analysis of MRI has received

Soft-tissue tumors form a second major area of our interest. PhD student Douwe Spaanderman developed a novel minimally interactive method to accurately contour such tumours in a time-efficient way (see Figure 2). This method showed excellent generalizability for both CT and MRI scans, in 14 different types of soft-tissue tumors. Also, he performed a comprehensive external validation of a previously developed radiomics model classifying lipoma and well-differentiated liposarcoma on MRI. In 2025, thanks to a comprehensive joint effort by PhD students Xinyi Wan, Matthew Marzetti, and Douwe Spaanderman, we published a systematic literature review on AI in radiological imaging of soft-tissue and bone tumours, including 325 papers. Finally, in collaboration with Dr. Thomeer (see p. 178) and Dr. Starmans, a third major area of interest was launched in 2024: the Liver Artificial Intelligence (LAI)

scientific report 2025 | APPLIED MEDICAL IMAGE ANALYSIS

Figure 2: InteractiveNet - our interactive method for segmenting soft-tissue tumours.

consortium. The mission of the LAI consortium is to fastforward the development, validation, and implementation of machine learning methods that could support MRIbased diagnosis of liver lesions. A dedicated website was launched: https://lai-consortium.org/. Two PhD students started on this project: Frederik Hartmann and Ruben Niemantsverdriet. In 2025, major progress in data collection has been made, and the consortium has been presented at various conferences (ECR, ESGAR, EASL).

Health Data Science Infrastructure With our team of research software engineers led by Hakim Achterberg, and strong involvement of Dr. Bron and Dr. Starmans, we are actively participating in EUCanImage, EOSC4Cancer, SHAIPED, and EUCAIM: largescale international projects funded by the European Commission to build secure and federated infrastructure for next-generation artificial intelligence in oncology. On national level we are contributing strongly to Health-RI, managing the imaging community and jointly establishing an infrastructure for data management, sharing, analysis and reuse of medical imaging data. On a European level, we are hosting the Euro-BioImaging Population Imaging node, offering our tools and services in this domain.

Expectations & Directions In the upcoming years, we will further expand both the fundamental research on development of new image analysis algorithms, and the applied research, in which we test the newly developed methods in clinical applications. Moreover, we expect to continue our activities in building infrastructure to facilitate image data management, data sharing, and data re-use for research, both in national and international initiatives. Last but not least, we expect to intensify our contributions to education on AI in medicine, as highlighted on page 69 by Dr. Jifke Veenland.

Funding Lekadir, Karim (Universitat de Barcelona), Aad van der Lugt, Wiro Niessen, Stefan Klein, Daniel Bos, and consortium partners EU Horizon2020: ‘EUCanImage – A European Cancer Image Platform Linked to Biological and Health Data for Next-Generation Artificial Intelligence and Precision Medicine in Oncology’. 2020-2025 Klein, Stefan, Jan-Jaap Visser, Dirk Grunhagen, Kees Verhoef, Stefan Sleijfer, Wiro Niessen, Arno van Leenders, and Martijn Starmans Hanarth Fonds: ‘Automatic grading and phenotyping of soft-tissue tumors through machine

learning to guide personalized cancer treatment’. 20212025

Stefan Klein. ‘cBioPortal & Radiology’. cBioPortal symposium, Amsterdam, the Netherlands. Nov 2025.

Marti-Bonmati, Luis (La Fe Polytechnic and University Hospital), Esther Bron, Martijn Starmans, Marcel Koek, Jan-Jaap Visser, Wiro Niessen, Stefan Klein, and consortium partners EU HORIZON DIGITAL-2022-CLOUDAI-02: ‘EUCAIM: European Federation for Cancer Images’. 2023-2027

Stefan Klein. ‘AI for Medical Image Analysis’. 15 years Predict, Rotterdam, the Netherlands. Nov 2025.

Niessen, Wiro, Aad van der Lugt, Stefan Klein, Juan A Hernandez-Tamames, Edwin Oei, Dirk Poot, Marion Smits, Jan-Jaap Visser, and consortium partners NWO, Min EZK and GE Healthcare: ‘ROBUST consortium: Trustworthy AI for MRI ICAI lab’. 2022-2027 Rousian, Melek Stichting Sophia Wetenschappelijk Onderzoek (SSWO): ‘A deep learning solution for congenital anomaly detection in early pregnancy’. 2022-2026 Hirvasniemi, Jukka, Stefan Klein, and Edwin Oei (coapplicants), Sita Bierma-Zeinstra (General Practice), and Jaap Harlaar (TU Delft) (main applicants) TU Delft-Erasmus MC Convergence Flagship: ‘Healthy Joints’. 2022-2027 Klein, Stefan, Marcel Koek, Hakim Achterberg, Adriaan Versteeg, Martijn Starmans, and consortium partners ISIDORe JRA PROGRAMME: ‘PATH2XNAT: Covid19 Pathomics meets XNAT’. 2023-2025

Wietske Bastiaansen. ‘AI for Prenatal Image Analysis (AI4PIA)’. Technical University Eindhoven, the Netherlands. Feb 2025. Wietske Bastiaansen. ‘AI for Imaging Processing’. Clinical learning evening Rotterdam area gynecology specialist training, Sint Fransiscus Gasthuis hospital, Rotterdam, the Netherlands. Jan 2025. Wietske Bastiaansen. ‘Clinician meets technician: what will be possible in the future’. Opening Outpatient Center for Innovative Fetal Imaging, Erasmus MC, Rotterdam, the Netherlands. Jan 2025.

Additional Personnel Laura Estacio Cerquin – PhD student at Netherlands Cancer Institute Abdullah Thabit – Research Software Engineer

Klein, Stefan, Maarten Thomeer, and Martijn Starmans NWO Open Technology Programme: ‘The Liver Artificial Intelligence (LAI) consortium: a benchmark dataset and optimized machine learning methods for MRI-based diagnosis of solid appearing liver lesions’. 2024-2028 Rousian, Melek, Wietske Bastiaansen, Stefan Klein, Régine Steegers-Theunissen, Martin Mienkina, and Stephan Anzengruber Erasmus MC-TKI-LSH Health Holland: ‘Towards Fully automated Anomaly Screening in the first Trimester of pregnancy using Artificial Intelligence (FASTAI)’. 2024-2028 Išgum, Ivana, Clarisa Sánchez, Marius Staring, Rob van der Geest, Keelin Murphy, Mitko Veta, Josien Pluim, Stefan Klein, and Mathias Prokop NWO Perspectief: ‘AI4AI: Artificial Intelligence for Accessible Medical Imaging’. 2024-2029

Invited Lectures Stefan Klein. ‘EUCanImage Metadata Catalog’. European Congress of Radiology (ECR), Vienna, Austria. Feb 2025. Stefan Klein. ‘Health-RI Imaging Community: ontwikkeling en validatie van AI voor medische beeldanalyse op data uit heel Nederland’. Data Driven Healthcare, Rotterdam, the Netherlands. June 2025. Stefan Klein. ‘EUCanImage Data Platform and Metadata Catalog’. European Association for Cancer Research, Online. Sept 2025. Stefan Klein. ‘Artificiële Intelligentie (AI) voor de Zorg’. Contactdag GIST, Amersfoort, the Netherlands. Sept 2025.

scientific report 2025 | APPLIED MEDICAL IMAGE ANALYSIS

Associate Professor

Jifke Veenland, PhD Email j.veenland@erasmusmc.nl

Jifke Veenland has a MSc degree both in medicine and in computer science, and a PhD in medical image processing and pattern recognition. The focus of her research is on tissue characterization and quantification of heterogeneity in MRI tumor images by applying Machine Learning and/or Deep Learning. The main part of her activities are in the field of education: she is coordinator of the MSc track Imaging & Interventions of Technical Medicine, the joint degree program of the Erasmus MC, LUMC and the TU Delft. Next to that she is project leader of the ErasmusArts 2030 Technology group to develop a curriculum for digitally skilled doctors. She is coordinator of the modules Imaging and Image Processing for the BSc Clinical Technology and for the MSc Technical Medicine. She coordinates the modules Advanced Image Processing and Machine Learning.

Machine and Deep Learning

Blastocyst Segmentation in IVF

Machine and deep learning techniques are increasingly used for medical imaging analysis. These methods help segment lesions, such as tumors in prostate MRIs, and predict treatment outcomes, like chemotherapy response in breast cancer patients. Training technical medicine students, data scientists, and AI-smart doctors is crucial for advancing this field.

(In collaboration with the department of gynaecology) IVF treatment involves selecting viable embryos based on their growth in the EmbryoscopeTM. We developed a deep learning model to segment blastocysts from timelapse images, generating growth curves. By combining these with clinical features, a machine learning model achieved a ROC-AUC of 0.70 this can improve embryo selection and IVF success rates.

Research Projects

Education Projects

Prostate Cancer Segmentation

Digital Technology in Medical Education

(In collaboration with the department of pathology) Convolutional Neural Networks (CNNs) were developed to detect and segment clinically significant prostate cancer in MR images. A multi-class segmentation model was trained to classify different tumor aggressiveness levels, which can guide urologists in biopsy procedures.

With the 2024 renewal of the Erasmus MC BSc Medicine curriculum, digital literacy is a key focus. Medical students are introduced to AI, e-health, VR/AR, and smart predictive algorithms. Learning goals and materials have been developed to train future doctors in identifying opportunities and ethical challenges in digital healthcare.

Breast Cancer Outcome Prediction (In collaboration with the department of pathology)

Participation in the SUSA Project

Triple-negative breast cancer (TNBC) patients often have poorer outcomes. Neoadjuvant chemotherapy (NAC) reduces tumor size, but not all patients respond effectively. Using deep learning models on DCE-MRIs at two time points, we aim to predict pathologic complete response, improving personalized.

The Sustainable Healthcare with Digital Health Data Competence (SUSA) project, funded by the EU Digital Europe Programme, enhances digital competencies in healthcare. Integrating digital skills into 20 bachelor’s and 26 master’s programs, along with lifelong learning modules, SUSA aims to train over 6,500 graduates and upskill 600 professionals, aligning with EU goals for sustainable and data-driven healthcare.

Post-doc

Wietske Bastiaansen, PhD AI for Prenatal Image Analysis

Every child deserves the best possible start in life, and proper prenatal growth monitoring is essential. Impaired growth, such as fetal growth restriction (FGR) or small-for-gestational-age (SGA) infants, raises perinatal risks and lifelong health challenges. Parental health and lifestyle factors further impact development, but 2D ultrasonography, the current standard, is limited in assessing these influences. While 3D ultrasonography offers richer insights, it demands expert training and is time-intensive. To address this, I focus on AI-basedmethods to automate 3D ultrasound analysis of the embryo, fetus, and placenta.

PhD Students Frederik Hartmann, MSc

Gonzalo Mosquera Rojas, MSc

Advisors Martijn Starmans, Maarten Thomeer & Stefan Klein

Advisors Stefan Klein & Marion Smits

The Liver Artificial Intelligence (LAI) Consortium Differentiating solid liver lesions on MRI is challenging due to their variety, different protocols, and rare subtypes. Accurate methods are needed to guide care and treatment. The LAI consortium aims to develop and implement machine learning tools for MRI-based diagnosis of liver lesions while promoting open science through public data sharing. Twelve hospitals worldwide are collecting MRI scans and clinical data from over 3,000 patients to support this effort.

Trustworthy AI for glioma diagnosis This project focuses on the development of trustworthy AI models for glioma diagnosis using MR imaging, which has the potential to improve the treatment decision making process. The design of the models has a strong focus on explainability and uncertainty quantification, aiming to ensure interpretable results in clinical settings.

Nikolai Herrmann, MSc Advisors Stefan Klein, Régine Steegers Theunissen, Wietske Bastiaansen & Melek Rousian

Ruben Niemantsverdriet, MSc Advisors Stefan Klein, Martijn Starmans & Maarten Thomeer

Novel, robust and resource-efficient volumetric biomarkers for first-trimester 3D ultrasound during pregnancy

The Liver Artificial Intelligence (LAI) Consortium

To develop failure-aware and resource efficient AI methods for automated embryonic growth monitoring systems in three-dimensional (3D) ultrasound. This includes automatically performing embryonic bodypart segmentations and from these, derive quantitative biomarkers for monitoring development. A strong focus is put on explainability and efficiency to facilitate implementation in clinical practice.

Differentiating solid liver lesions on MRI is challenging due to their variety, different protocols, and rare subtypes. Accurate methods are needed to guide care and treatment. The LAI consortium aims to develop and implement machine learning tools for MRI-based diagnosis of liver lesions while promoting open science through public data sharing. Twelve hospitals worldwide are collecting MRI scans and clinical data from over 3,000 patients to support this effort.

scientific report 2025 | APPLIED MEDICAL IMAGE ANALYSIS

Douwe Spaanderman, MSc Advisors Stefan Klein, Martijn Starmans, Dirk Grünhagen & Wiro Niessen AI-driven Soft Tissue Tumor Management Soft tissue tumors (STTs) represent a rare and diverse group of lesions characterized by a wide spectrum of differentiation. These subtypes vary significantly in clinical behavior, aggressiveness, molecular profiles, and optimal treatment strategies. To address this complexity, we are leveraging advanced machine learning techniques to analyze medical imaging, such as computed tomography (CT) and magnetic resonance imaging (MRI), for improved tumor detection and prognosis prediction.

Marcella Zijta, MSc Advisors Stefan Klein, Melek Rousian, Bernadette de Bakker & Wietske Bastiaansen Congenital Anomaly Detection In Early Pregnancy The project focuses on developing AI-driven methods for anomaly detection, organ segmentation, and growth assessment in embryos and fetuses using 3D ultrasound and Micro-CT scans. Key objectives include estimating developmental age, detecting congenital anomalies, and automating first-trimester anomaly screening to enhance early diagnosis and monitoring.

Dirk Poot, PhD assistant professor 72

QUANTITATIVE MRI RECONSTRUCTION

Top Publications 2025

Dirk Poot is asssistant professor and heading the quantitative MRI reconstruction research line. He is affiliated with the Biomedical Imaging Group Rotterdam (BIGR, http://www.bigr.nl) and the MR physics group. In 2005, Dirk received his MSc degree from the faculty of Applied Physics at the Delft University of Technology, Delft/NL. In 2010 he obtained his PhD degree at the Visionlab, University of Antwerp, Antwerp/BE, for his research on reconstruction and statistical processing of Magnetic Resonance Images. He is lab manager of the new ICAI lab Trustworthy AI for Magnetic Resonance Imaging. His current research interests include MR image acquisition, reconstruction, quantification, and motion compensation.

Byanju R, S Klein, A Cristobal-Huerta, JA HernandezTamames, DHJ Poot. Myelin water imaging from accelerated 3D-GRASE acquisitions using subspace constrained reconstruction, MAGMA 38 2025; 933948. van Dorth D, K Venugopal, KN van Der Werff, M Smits, EAH Warnert, JA Hernandez-Tamames, MJP van Osch, DHJ Poot. Exploring the need for a preload on the estimation of permeability, vessel radius, and relative cerebral blood volume in MR vascular fingerprinting-based dynamic susceptibility contrast perfusion imaging, Magnetic Resonance in Medicine. 2025; 93: 1761-1770.

d.poot@erasmusmc.nl Fuderer M, B Wichtmann, F Crameri, NM de Souza, B Baessler, V Gulani, M Wang, DHJ Poot, R de Boer, M Cashmore, KE Keenan, D Ma, C Pirkl, N Sollmann, S Weingartner, S Mandija, X Golay. Colormap recommendation for MR relaxometry maps. Magnetic Resonance in Medicine 2025; 93: 490-506.

Context

uantiative MRI is becoming increasingly relevant in the era of precision medicine. Current clinically used MRI protocols are still mostly limited to weighted images, such as Tl-weighted or T2-weighted. This delivers images optimized for visual inspection by a radiologist that is looking for structural abnormalities. However, these images do not provide measurements of the actual magnetic resonance properties of the tissue; e.g. the Tl or T2 relaxation time, or the diffusion or perfusion rate. Also, there might be substantial variability in the images between scanners, or even from the same scanner at different moments in time. This lack of standardization hampers the detection of subtle diffuse, disease induced, changes in the tissues. The key objective of quantitative MRI is to complement the qualitative images with quantitative measurements of tissue properties. We have a substantial number of projects in which different properties are measured.

Research Projects: Objectives & Achievements

Vascular properties For tumor growth vascularity Is highly relevant. With the development of novel acquisition and analysis methods we aim to extract more information on the vasculature, which we aim to develop into biomarkers.

My research line focusses on quantitative MRI reconstruction, in close collaboration with the MR physics group of J.A. Hernandez-Tamames (page 34) as well as the image registration group of S. Klein (page 60). The aim of quantitative MRI is to objectively measure tissue properties such as for example the T1, T2(*) relaxation times, temperature or tissue perfusion. Traditionally, this is done by acquiring several images with specific differences in their acquisition settings such as echo time or inversion. The intensity of the acquired images is fitted to a model that is derived from the MR physics of the acquisition method. Current projects: – Diffusion model for qMRI (Shishuai Wang) – Resolution enhancement of qMRI by using weighted images (Alireza Samdifardheris) – Vascular signature mapping (Karen van der Werff) – qMRI reconstruction with motion compensation (Bram van Audekerke) – DL-based qMRI for multi-compartment diffusion (Guanqun Liu)

Expectations & Directions In the upcoming years, we aim to further develop the quantitative MR image acquisition and reconstruction methods. In parallel the novel developments will be applied in clinical research projects for further evaluation and in collaboration with industry we aim to bring the developments to actual clinical use.

Funding Poot, Dirk, Juan Hernandez Tamames, Stefan Klein, Aad van der Lugt, Edwin Oei, Marion Smits, Jan-Jaap Visser, and consortium partners NWO, Min EZK and GE HealthCare: ‘ROBUST consortium: Trustworthy AI for MRI ICAI lab’. 2022-2027 Poot, Dirk, Juan Hernandez Tamames, Stefan Klein, and consortium partners Marie Curie: 'IQ-BRAIN'. 2024-2028

Acceleration of acquisition Within the new Trustworthy AI ICAI lab one of the projects is to further accelerate the image acquisition by severely under-sampling the k-spaces, such that normal image reconstruction fails. By using the known MR physics as well as deep learning approaches we aim to still obtain highquality images and tissue property maps from the short scans. The trustworthiness of these resulting images to correctly reflect the specific patient is key in this research. (Fig 1.)

Hernandez Tamames, Juan, Dirk Poot and Frans Vos TKI Health~Holland and GE Healthcare: 'Gadolinium free enhancement with magnetic resonance imaging synthesis'. 2024-2028

Invited Lectures Dirk Poot. ‘Quantitative MR reconstruction’, MSK Symposium, Rotterdam, the Netherlands. March 2025.

Motion compensation Subject motion is a major cause of low quality or failed MRI exams. For quantitative MRI acquisitions this situation is even worse, as typically they are longer than the acquisition of individual traditional weighted images. Additionally, in the analysis, images are combined, increasing the sensitivity to motion. Hence a major focus of my research line is to compensate for subject motion. By exploiting the known relations among the images as well as by adjusting the acquisition to acquire some reference data to allow identification and subsequent compensation of unavoidable subject motion.

Acceleration of reconstruction

The advanced methods that we develop for high quality reconstruction from highly accelerated scans may have long computation times. Even though computers get faster every year, innovative methods to improve computation time are needed. With novel work on deep learning based image reconstruction we are accelerating the reconstruction process to achieve clinically acceptable reconstruction times for the advanced methods.

Figure 1. From Dorth et al, MRM 93, figure 5. This figure shows parametric maps for permeability (A,B), radius (C,D), and relative cerebral blood volume (rCBV; E,F) for the first (A,C,E) and second (B,D,F) bolus of a patient. The first bolus permeability map clearly shows less over-estimation of the expected zero permeability in normal brain tissues.

scientific report 2025 | QUANTITATIVE MRI RECONSTRUCTION

PhD Students Bram van Audekerke, MSc

Guanqun Liu, MSc

Advisors

Stefan Klein, Juan Hernandez Tamames & Dirk Poot

Juan Hernandez-Tamames, Stefan Klein & Dirk Poot

QMRI reconstruction framework with motion compensation, uncertainty estimation and segmentation

Deep learning-based multiparametric diffusionrelaxometry and data harmonization across scanners and sites

Bram’s project is focused on implementing deep learning for the reconstruction of quantitative MR maps from undersampled MR acquisitions. Bram is currently investigating a multiparametric reconstruction framework and retrospective motion compensation with future plans of estimating uncertainty on acquired quantitative values.

Guanqun’s project focuses on developing deep learning methods for multiparametric diffusionrelaxometry, aiming to reconstruct quantitative MRI measures robustly and efficiently. His current work is to improve data harmonization of quantitative maps across scanners and sites, enabling more consistent and trustworthy measurements for research and clinical applications.

Emanoel Sabidussi, MSc

Alireza Samadifardheris, MSc

Advisors Stefan Klein & Dirk Poot

Advisors Juan Hernandez-Tamames, Stefan Klein & Dirk Poot

Advanced deep learning methods for Quantitative MRI

Trustworthy AI for adaptive and precision MR protocols

With Recurrent inference machines Emanoel could quantify relaxometry as well as diffusion properties. With this model based technique the strengths of the knowledge of the MR physics are combined with the power of deep learning in a way that allows training the deep learning method with simulated data. State of the art results are obtained.

In the first project, Alireza applies deep learning to superresolve low-resolution MR quantitative maps using high-resolution weighted images. This is the first step towards obtaining fast multiparametric maps for abnormality detection and subsequent online protocol adjustment.

Karen van der Werff, MSc

Shishuai Wang, MSc

Advisors

Marion Smits, Stefan Klein, Frans Vos & Dirk Poot

Stefan Klein, Juan HernandezTamames & Dirk Poot

Brain tumor genotyping based on multicomponent vascular fingerprinting

End-to-end deep learning quantitative MR reconstruction

Karen is developing a MR fingerprinting technique for multi-component estimation of vascular properties. This will result in quantitative maps for different vascular parameters. The performance of these maps in terms of glioma genotyping will be analyzed, with the aim of non-invasively characterizing brain tumours.

Shishuai developed deep-learning based methods to directly reconstruct multiple high-quality quantitative MR maps (e.g. T1, T2 and proton density) from highly under sampled k-space data acquired with a Quantitative Transient Imaging protocol. Shishuai is also investigating quantitative MR mapping by using diffusion model.

Esther E Bron, PhD associate professor 76

NEUROIMAGE ANALYSIS & MACHINE LEARNING

Top Publications 2025

Esther Bron is Associate Professor in Medical Image Analysis and is heading the Neuroimage Analysis and Machine Learning research line. She is affiliated with the Biomedical Imaging Group Rotterdam (BIGR, http://www.bigr.nl). In 2011, Esther received her MSc degree in Medical Natural Sciences – specialization Medical Physics, cum laude – at the VU University, Amsterdam/NL. Esther successfully defended her doctoral thesis in 2016 at the Erasmus University Rotterdam, which focused on the development and validation of advanced image analysis techniques for computeraided diagnosis of dementia. Since then, she has successfully built her research line and has been appointed associate professor in 2025. Esther has received several (young investigator) awards and is management board member in several (inter) national consortia (e.g. TAP-Dementia, EUCAIM). In addition, she leads the imaging working group of Health-RI, the national health data infrastructure in the Netherlands.

Mateus P, S Garst, J Yu, D Cats, AG Harms, M Birha nu, M Beekman, PE Slagboom, M Reinders, J van der Grond, A Dekker, JF Jansen, M Beran, MT Schram, PJ Visser, J Moonen, M Ghanbari, G Roshchupkin, D Vojinovic, I Bermejo, H Mei, EE Bron. Multi-Cohort Federated Learning Shows Synergy in Mortality Prediction for MRI-Based and Metabolomics-Based Age scores. Journal of Healthcare Informatics Re search 2025; 9:606-628. Liu X, B Li, MW Vernooij, EB Wolvius, GV Roshchupkin, EE Bron. AI-based association analysis for medical imaging using latent-space geometric confounder correction. Medical Image Analysis 2025; 102:103529. Martí-Bonmatí L, I Blanquer, M Tsiknakis, G Tsakou, R Martinez, S Capella-Gutierrez, S Zullino, J Meszaros, EE Bron, JL Gelpi, K Riklund, L Chaabane, HP Schlemmer, M Aznar, P Serrano Candelas, P Gordebeke, M Hierath. The EUCAIM Consortium, Empowering Cancer Research in Europe: The EUCAIM Cancer Imaging Infrastructure. Insights into Imaging 2025; 16:47.

e.bron@erasmusmc.nl

Context

rain diseases including dementia impose an enormous burden to the individual and to society. As a consequence, there is an urgent need to develop effective preventive and therapeutic strategies, especially in dementia. Early detection and accurate prediction of the progression of at-risk subjects are key in this development. Artificial Intelligence (AI) has high potential for aiding medical decision making and creating a learning healthcare system, where current patients are diagnosed and treated based on knowledge gained from previous patients. Especially in the domain of neuroradiology, AI has proven very successful in gaining new knowledge by extracting patterns related to neurodegenerative diseases from large sets of MRI data. My mission is to translate AI to clinical practice, so future patients can be diagnosed and treated based on knowledge gained from previous patients.

Research Projects: Objectives & Achievements

With the Imaging Office team, we have worked on major contributions to the Cancer Image Europe (EUCAIM) platform (fig. 2), in which we contribute as WP lead, imaging data reference node and provide expertise and development on the imaging data catalogue and metadata model.

AI for image-based diagnosis and prediction I develop with my team novel image analysis and prediction methodology for neurodegenerative and cardiovascular diseases and strongly collaborate with clinicians and clinical researchers to validate these methods for clinical application. This entails AI for automatic analysis of multimodal MRI brain scans, including perfusion MRI and vascular imaging biomarkers. We developed a novel disease progression model and evaluated this in several neurodegenerative diseases (i.e. Alzheimer’s disease, frontotemporal dementia). Current work focuses on developing diagnostic AI that can be translated to clinical practice and its validation on large-scale clinical data. Deep learning methods have large potential for early detection in dementia. However, there are some limitations of deep learning methods that have prevented them from being routinely used in clinical practice. A key problem is the limited interpretability of the predicted results, because deep learning models often operate as a black box. Wenjie Kang published his journal paper on the Glo&Loc-EBM method (Fig. 1). This novel method combines the interpretability of Explainable Boosting Machines with deep learning techniques exploiting the high-dimensional information captured by imaging. Another challenge in dementia diagnosis is the etiological diagnosis of different diseases underlying dementia. Myrthe van Haaften works on several projects related to this. Based on the initial work of three MSc students, she used hierarchical clustering to improve understanding of imaging patterns in memory clinical patients related to their diagnosis (Van Haaften et al, 2026).

Figure 1. The novel method Glo&Loc-EBM (red) selects relevant regions based on an occlusion map, computes deep learning features and performs classification based on an explainable boosting machine (EBM). The method achieves higher performance for classifying Alzheimer’s disease patients from controls than previous methods (Kang at al., Computerized Medical Imaging and Graphics, 2026).

Current projects: - Explainable machine learning methods (Wenjie Kang, Bo Li) - AI for etiological diagnosis of dementia (Myrthe van Haaften, Kaouther Mouheb) - Subtypes of healthy aging and early detection of cognitive decline (Sterre de Jonge, Bailey Berman) - Predictive modeling in cardiovascular disease (Hyunho Mo, Ashley Mangrey) - AI for diagnosis of FTD pathology (Laura Dias, Viktoria Ehret) - Disease onset prediction in genetic FTD (Wenjie Kang, Myrthe van Haaften, Laura Dias, Jon Bregman) - Image analysis of morphology in craniosynostosis (Diede Wijnbergen, Henri Vrooman)

Figure 2. Architecture of the Cancer Image Europe infrastructure (Marti-Bonmarti et al., 2025)

scientific report 2025 | NEUROIMAGE ANALYSIS & MACHINE LEARNING

Imaging data accessibility

Expectations & Directions

Medical AI development needs large and heterogeneous datasets. A key bottleneck here is clinical data sharing, which is difficult -or sometimes impossible- because of privacy and safety issues. The novel technology of federated learning provides a new angle to solving this universal problem. Opposed to bringing all data to a central place for training a machine learning algorithm (central learning), federated learning works with several computer nodes that train an algorithm collaboratively without exchanging the data. In biomedical imaging, federated learning has recently been applied to multiple research tasks, including whole-brain segmentation and brain-age prediction. While the technique cannot be applied in a routine way to clinical data yet, I aim with my team to set a major step in overcoming methodological (e.g. different data distributions, quality control) and organisational (e.g. privacy and security, workflow) challenges. I consider federated analyses to be essential for the success of AI in medicine and together with my team we have the expertise to boost this.

In coming years we continue to study novel artificial intelligence (AI) methods for image analysis and outcome prediction in the field of neurological and cardiovascular diseases with a focus on dementia, addressing challenges related to clinical translation of these methods. In 2025, new projects will start (CHIME, Predict-FTD). In addition, through national and European research infrastructure initiatives, we work on pushing the field forward in making imaging and health data accessible for research and innovation in an efficient yet secure manner.

Funding Van der Harst, Pim, Folkert Asselbergs, Michiel Rienstra, Lotte Krabbenborg, Wiro Niessen, Daniel Bos, and Esther Bron NWO Commit2Data Big data & health grant: ‘MyDigiTwin: Early recognition and prevention of cardiovascular diseases’. 2022-2026 Kang, Wenjie, and Esther Bron China Scholarship Counsil (CSC) Fellowship: ‘Interpretable machine learning for diagnosis and prognosis in neurodegenerative disease’. 2022-2026

Related to this, deep learning are known to be susceptible to biases in the data. Especially, biases against specific groups may hinder clinical applicability. A recently discovered phenomenon, Neural Collapse (NC), has shown potential in improving the generalization of state-of-theart deep learning models. Nonetheless, its implications on bias in medical imaging remain unexplored. Kaouther Mouheb presented at the 2024 MICCAI conference a study that investigated deep learning fairness through the lens of neural collapse and analyzed its training dynamics. We found that biased training initially results in different neural collapse configurations across subgroups, before converging to a final solution by memorizing all data samples (Mouheb et al, 2024). I work as Image Data Coordinator at the Dutch National Research Infrastructure: Health-RI. Here, I lead (together with Dr. Stefan Klein) the Health-RI Imaging Working group. In 2024, key milestones were the onboarding of the first 5 imaging datasets in the National Healthdata Catalogue, an associated software library (Img2catalog) and an advisory document to the Health-RI nodes on the recommended approach for onboarding imaging data. In addition, we wrote an document for Health-RI to advise general hospitals that aim to participate in multi-center studies on using CTP as software for image data de-identification. Current projects: – Federated learning for imaging data (Kaouther Mouheb, Hyunho Mo, Alexander Harms, Mahlet Birhanu) – Large-language models (LLMs) for extracting neuroradiology ratings (Kaouther Mouheb, Antoine Manenti) – Meta-data models for imaging data (Alexander Harms, Mahlet Birhanu, Hakim Achterberg) – Cancer Imaging Europe dashboard and catalogue (Alexander Harms, Mahlet Birhanu, Hakim Achterberg) – Imaging Office / Euro-BioImaging Population Node (Ilva van Houwelingen)

Bron, Esther Erasmus MC Fellowship grant: ‘Etiological diagnosis of dementia using federated artificial Intelligence’. 2023-2027 Vernooij, Meike, Esther Bron, Hugo Kuijf, Geert Jan Biessels, Vikram Venkatraghavan, and Betty Tijms ZonMW TAPDementia grant: ‘Timely Accurate Personalized Diagnosis using Artificial iNtelligence to Classify dementia Etiologies (TAP-DANCE)’. 2023-2027 Marti-Bonmati, Luis, .. , Wiro Niessen, Esther Bron, Stefan Klein, and Martijn Starmans Digital Europe Programme (Horizon Europe) grant: ‘EUropean Federation for CAncer IMages (EUCAIM)’. 2023-2027 Klomp, Dennis, Mariëlle Emmelot-Vonk, Geert Jan Biessels, Esther Bron, and Meike Vernooij Health~Holland LSHmatch multicenter PPP: ‘Scan2go: autonomous MRI for large scale diagnostics of brain integrity’. 2023-2027 Zwanenburg, Jacobus, Natalia Petridou, Thijs van Osch, Geert Jan Biessels, Esther Bron, Meike Vernooij, et al. NWO Applied and Engineering Sciences Perspectief Grant: 'CHIME: Cerebral HemodynamIcs, Metabolism and clearancE: A comprehensive, non-invasive brain imaging approach to characterize key biological processes in dementia'. 2024-2028 Vernooij, Meike, Esther Bron, Frank Wolters, Eline Vinke, Jeroen de Bresser, Frans Vos, Thijs van Osch, Simon Mooijaart, and Harro Seelaar Medical Delta: 'Medical Delta Dementia 3.0: Applying advanced brain imaging for efficient dementia diagnosis and prediction: when, what, and for whom?'. 2024-2028

Seelaar, Harro, Esther Bron, Barbare Borroni, Ruta Gostautaite, Gianluigi Zanusso, Raquel Sanchez, Anja Schneider, Eino Solje, Mireia Jofre-Bonet, Jonas Meirer, Angela Bradshaw, Markus Otto, Yolande Pijnenburg, Evelyn Moreno, and Alessio Fasano Horizon Europe ( H O R I Z O N - H LT H 2 0 2 4 - D I S E A S E - 0 3 - t w o - s t a g e ) : 'PREDICTFTD: Accelerating the Validation of Predictive Liquid Biomarkers for Frontotemporal Dementia Diagnosis and Subclassification'. 2024-2028

Additional Personnel Viktoria Ehret, MSc - Imaging Data Steward Jessica van Wijngaarden, MSc - Imaging Data Steward Hakim Achterberg, PhD - Research Software Engineer Adriaan Versteeg, MSc - Research Software Engineer

Invited Lectures

Ilva van Houwelingen, MSc - Process Coordinator Imaging Office

Kaouther Mouheb. ‘Trustworthy and Federated Learning with Medical and Imaging Applications’. Trustworthy and Federated Learning with Medical and Imaging Applications, Leuven, Belgium. Nov 2025. Esther Bron. ‘Towards Interoperable Data Spaces’. Health RI congerence, Utrecht, the Netherlands. Nov 2025. Esther Bron. ‘Sharing images: The why and how of medical imaging analysis research’. Data stewards course - HealthRI. Utrecht, the Netherlands. June 2025. Esther Bron and Stefan Klein. ‘Development and validation of AI for medical image analysis using data from across the Netherlands’. Data-Driven Healthcare. Healthcare, Rotterdam, the Netherlands. June 2025. Esther Bron. ‘AI in Radiology’. Graduate school MODEM. Rotterdam, the Netherlands. May 2025. Esther Bron. ‘MRI-based machine learning for diagnosis and prediction in dementia’. Translational Neuroimaging Meeting, Utrecht, the Netherlands. March 2025. Esther Bron. ‘Panel session: How to ensure availability of high-quality cancer imaging data to support AI-based innovation?’. European Commission - European Cancer Imaging Initiative. Brussels, Belgium. Feb 2025. Esther Bron. ‘Image Analysis Methodologies’. European School of Radiology (ESOR): Fundamentals of Radiological research. Feb 2025.

Jon Bregman, MSc - Process Coordinator Imaging Office Ashley R. Mangrey - BSc Student Bailey Berman - MSc Student

scientific report 2025 | NEUROIMAGE ANALYSIS & MACHINE LEARNING

Assistant Professor

Henri Vrooman, PhD Email h.vrooman@erasmusmc.nl

Henri A. Vrooman received the Ph.D. degree in physics (1991) from Delft University of Technology (TU-Delft), the Netherlands. From 1986 to 1990, he was a Research Scientist at the Department of Image Processing and Pattern Recognition of the TU-Delft. From 1990 until 2000 he was involved in several image processing projects in cooperation with the Department of Radiology at the Laboratory for Clinical and Experimental Image Processing of Leiden University Medical Center. Since April 1st, 2000, Henri is Assistant Professor at the Erasmus MC - University Medical Center - Rotterdam, the Netherlands, where he initiated the Biomedical Imaging Group Rotterdam (BIGR). His research interests include digital image processing, pattern recognition, biomedical image processing techniques and diagnostic radiology. The recent focus of his work is on neuroimaging and on the development of infrastructures for the processing of large data sets, especially important for Population Imaging.

Providing image processing as a service In 2025, the Imaging Office, embedded in the Department of Radiology & Nuclear Medicine, received more and more requests for processing and annotation tasks. The office targets parties that need access to medical imaging data and analysis tools. This includes support for image acquisition, data storage and analysis of imaging data with several pipelines. Examples of services we provided in 2025 were the processing of brain MRI images of children with rare diseases to extract relevant volumetric biomarkers, the extension of an annotation tool to perform 3D measurements of the pineal and annotate present cysts, and the automated detection of white matter lesions, using an advanced neural network approach.

Advanced processing in craniosynostosis Another interesting project we are Involved in for years now, is the processing of images from children with craniosynostosis. Metopic synostosis patients are at risk for neurodevelopmental disorders despite a negligible risk of intracranial hypertension. To understand underlying pathophysiology, we extended our portfolio with more advanced processing techniques, such as brain tissue segmentation, computation of the local gyrification Index (LGI), and arterial spin-labeling to measure regional cerebral blood flow. For tissue segmentations, scans were processed with Infant FreeSurfer. The scans of children younger than one year differ from adult imaging data. By giving age as a parameter, a different brain atlas is used to compute the Desikan cortex parcellation. The LGI of the

Figure 1. Processing of brain images from patients with nongenetic metopic craniosynostoses. The native scan of a patient (top-left), the Desikan cortex parcellation (top-right), and the surfaces for computing the LGI (bottom-left) are shown. At the bottom right, an example of segmented ventricles of another patient.

cortex is computed by smoothing the pial surface and taking the local distances between the two surfaces. Finally, we were working on methodology to segment brain ventricles In MRI scans. Metopic patients may have strangely shaped ventricles with an increased size. Standard processing techniques are unusable. We are working on a neural network approach to detect and properly segment the diseased ventricles. So far, the training set Is based on manual segmentations but will be extended with automatically segmented ones.

Post-docs

Bo Li, PhD Translatable AI in Neuroimage Analysis Information technology, such as Artificial Intelligence (AI), has the potential to transform the field of radiology by improving diagnostic accuracy, increasing efficiency and reducing healthcare costs. While in order to integrate AI tools into clinical workflows, it is critical that we design and validate them to be not only accurate, but also effective, unbiased, and transparent. Therefore, in my role of developing AI algorithms for brain MRI analysis, we focused on: 1.

simplifying and effectively supporting the imaging assessment of cerebral lacunes by additionally predicting the presence and a burden score of the lesion, besides normal segmentations (2nd place at VALDO challenge in MIC-CAI 2021; OHBM 2022); constructing neuroimaging endophenotypes for population studies that enables unsupervised dimensionality reduction at high-resolution, and allows for confounder controlling during feature construction (OHBM 2022, oral); and extending our novel AI algorithms for group-wise mean-space image registration to handle subject-wise longitudinal brain MRIs with gliomas, which is shown to serve as a fast (CPU run time: 0.1 mins) alternative to the stateof-the-art conventional toolboxes, such as ANTs (26 hrs), NiftyReg (23 mins), and Elastix (33 mins) (ISMRM 2023).

Hyunho Mo, PhD Cardiovascular Disease Risk Prediction using Deep Learning Algorithms Cardiovascular disease (CVD) is a leading cause of morbidity and mortality worldwide. Prevention of CVD requires accurate identification of high-risk individuals. Various CVD risk prediction models have been developed to identify persons at increased risk of CVD. However, those are commonly based on traditional linear regression models that are limited to using a small number of well-established risk factors typically assuming that there is no interaction between the different factors and each factor has a linear relationship with the CVD outcome. To this end, we are developing deep learning (DL) based CVD risk prediction models that can leverage associations between a larger number of predictors and their non-linear effects on the outcome. The use of DL algorithms is advantageous for developing end-to-end prediction models on datasets comprising medical imaging data as well as structured data for large populations. The datasets used in our work are made up of anthropometric measures, some risk factors, and cardiac/head CT scans that have associations with the risk of CVD. We build a DL-based framework that can process the multimodal data input. In this framework, we employ a convolutional neural network (CNN) to extract high-level image features that can be considered as predictors from imaging data. The CNN is followed by a DLbased CVD risk prediction model that can combine the obtained predictors (from 3D image) with tabular data to make CVD risk predictions. We aim to improve predictive accuracy in CVD risk prediction using DL algorithms.

scientific report 2025 | NEUROIMAGE ANALYSIS & MACHINE LEARNING

PhD Students Wenjie Kang, MSc

Myrthe van Haaften, MSc

Advisors Esther Bron, Bo Li & Stefan Klein

Advisors Esther Bron & Meike Vernooij

Interpretable deep learning for dementia

Deep learning for dementia diagnosis

Deep learning methods have large potential for early detection and prediction of dementia using highdimensional data such as neuroimaging. A key problem is the poor interpretability of the predicted results, because deep learning models often operate as a blackbox. In my PhD research, we aim to solve this problem by combining deep learning with interpretable machine learning methods to improve the explainability of deep learning model in the diagnosis and prediction of dementia.

Etiological diagnosis of dementia is a difficult task, complicated by disease heterogeneity within the etiologies and overlap between etiologies. Deep learning can provide datadriven insights in assigning the most probable etiology. In my PhD research, we will develop a deep learning model for etiological diagnosis of dementia based on multimodal input data, e.g. neuroimaging and neuropsychological exams.

Kaouther Mouheb, MSc Sterre de Jonge, MSc

Advisors Esther Bron & Stefan Klein

Advisors Esther Bron, Eline Vinke & Meike Vernooij

Federated learning for dementia diagnosis Diagnosing Dementia is challenging for clinicians. AI offers a potential solution, especially with the growing availability of radiological data. However, the sensitive nature of medical imaging data often prevents its sharing, presenting a significant obstacle for training effective AI models. Federated learning (FL) enables decentralized model training without data sharing. However, it faces challenges due to interclient data heterogeneity. In my PhD, I investigate how advanced AI techniques such as LLMs and foundation models can be integrated with federated learning to diagnose dementia more accurately and efficiently, while keeping patient data safe.

Early diagnostics in individuals with cognitive complaints using MRI & AI Early detection of dementia is challenging due to its preclinical phase, heterogeneity, and aging variations. Advances in AI and increasing availability of imaging provide new insights into brain patterns. I work on predicting dementia in individuals with cognitive complaints using disease progression models and AI, based on MRI brain markers.

Laura Dias, MSc Diede Wijnbergen, MSc

Advisors Esther Bron, Harro Seelaar & Meike Vernooij

Advisors Esther Bron, Irene Mathijssen & Tareq Abdel Alim

Development and validation of artificial intelligence (AI) tools based on imaging and liquid biomarkers for improving diagnosis of FTD pathology.

Imaging-based analysis of brain and craniofacial morphology in craniosynostosis Craniosynostosis is a congenital condition characterized by the premature fusion of one or multiple cranial sutures, leading to abnormal skull and brain development. Using advanced MRI analysis and 3D photogrammetry, the project aims to improve quantitative assessment of both brain morphology and craniofacial shape. The overall goal is to better understand morphological variation and development across craniosynostosis subtypes.

70-85% of patients with frontotemporal dementia (FTD) have a sporadic form of the disease. This diagnosis is slow (~3.6 years) with frequent misdiagnosis due to clinical, genetic and molecular heterogeneity. We aim to create AI models that assist the correct and timely diagnosis and subtyping of FTD based on imaging and liquid biomarkers.

Jukka Hirvasniemi, PhD assistant professor 84

MUSCULOSKELETAL IMAGE ANALYSIS

Top Publications 2025

JOINT APPOINTMENT IN TU DELFT (BIOMECHANICAL ENGINEERING)

Harlianto NI, J Hirvasniemi, DHJ Poot, S Klein, SMA Bierma-Zeinstra, D Schiphof, EHG Oei. T2 mapping of the articular cartilage as a biomarker for knee osteoarthritis: An analysis of the population-based Rotterdam Study. Osteoarthritis Cartilage 2025; 34:232-239.

Jukka Hirvasniemi is an assistant professor with a joint appointment in Biomechanical Engineering at TU Delft. He leads the research line in Musculoskeletal Image Analysis. Jukka obtained his MSc degree in Biomedical Engineering in 2011 and his PhD in 2015, both from the University of Oulu, Finland. He received a 3-year postdoctoral research grant from Academy of Finland in 2017 and conducted a one-year research visit to UMC Utrecht in 2018 funded by that grant. In 2018, he was awarded a 2-year Marie SkłodowskaCurie COFUND postdoctoral fellowship at Erasmus MC. He received a title of Docent (Adjunct Professor) in Medical Image Analysis from University of Oulu in 2019. Besides research, he is involved in various academic activities including lecturing for Technical Medicine students, editing and refereeing for various journals, and serving as a member of an international working group dedicated to advancing open and reproducible musculoskeletal imaging research. His research interests include development of artificial intelligence (AI) and advanced image analysis techniques to improve diagnosis and prediction of musculoskeletal diseases, such as osteoarthritis.

van Paassen R, N Tumer, J Hirvasniemi, TM Piscaer, AA Zadpoor, S Klein, SMA Bierma-Zeinstra, EHG Oei, M van Middelkoop. Patellar and Femoral Bone Morphology Is Associated With Overweight and Sports Participation in Young Adolescents. J Orthop Res. 2025; 43:2221-2232. Bonaretti S, M Barzegari, M Bevers, S Boyd, AJ Burghardt, D Cameron, F Chiumento, G Crimi, G Degenhart, P Durongbhan, MA Espinosa Hernandez, G Fraterrigo, A Ghasem-Zadeh, L Grassi, J Hirvasniemi, …, D Zukić. PREPRINT: Open and reproducible research in musculoskeletal imaging: Why it matters and how to implement it with the guidelines of the ORMIR community. Zenodo 2025; 10:ziag025.

j.hirvasniemi@erasmusmc.nl

Context

usculoskeletal disorders have a tremendous impact on the quality of life of an individual and impose a subtancial economic burden to society. Novel imaging, image processing, and image analysis techniques have a significant potential to enable earlier diagnosis and deeper understanding of musculoskeletal disorders and their pathophysiological processes. Additionally, AI techniques have great potential to enhance each component of the imaging chain. The Musculoskeletal Image Analysis research line focuses on using AI and advanced image analysis techniques to improve diagnosis and prediction of musculoskeletal diseases. This research line is part of the Biomedical Imaging Group Rotterdam (BIGR, https://www.bigr.nl/). We are closely collaborating with the ADMIRE research group (Prof. Oei) and Applied Medical Image Analysis research line (Prof. Klein). Our research projects include automatic segmentation of structures on MRI, imaging biomarker extraction from medical images, detection and diagnosis of abnormalities, and prediction of onset and progression of musculoskeletal diseases such as osteoarthritis.

Research Projects: Objectives & Achievements

Analysis of musculoskeletal imaging data from the Rotterdam Study

To reach our research goals, we are closely collaborating with the ADMIRE research group (PI: Edwin Oei) and Applied Medical Image Analysis research line (PI: Stefan Klein). We are also actively collaborating with researchers from the departments of General Practice and Orthopedics within Erasmus MC. External collaborations include TU Delft, UMC Utrecht, and University of Oulu.

In the Rotterdam Study cohort, we have developed deep learning-based automated MRI segmentation tools to assess bone and meniscus in the knee joint (Figure 2).

Analysis of musculoskeletal imaging data from the Generation R We are actively involved in the analysis of musculoskeletal imaging data from the Generation R and Rotterdam Study, two large population studies in Erasmus MC. In 2023, PhD student Mirthe Kamphuis started on the HIPSTAR project, investigating the impact of hip dysplasia on young adult joint integrity. She has developed automated AI tools for segmentation, radiomics, and morphology assessment of hip joints from MRI scans of the Generation R cohort (Figure 1). This project involves collaborations with departments of General Practice (Sita Bierma-Zeinstra, Jos Runhaar) and Orthopedics (Rintje Agricola, Jaap Tolk).

Figure 2. Automatically segmented MRI scan of a knee joint.

We have extracted radiomic features of bone from MRI to assess and predict osteoarthritis, T2 relaxation time values of articular cartilage to assess osteoarthritis, and meniscus volume to identify risk factors for osteoarthritis. Future analyses include extraction of 3D knee bone shape a potential biomarker for osteoarthritis. These projects are conducted in close collaboration with researchers from the Department of General Practice (Sita Bierma-Zeinstra, Jos Runhaar, Dieuwke Schiphof). We have also developed tools to automatically extract bone texture features from plain knee radiographs in the Rotterdam Study to assess osteoarthritis.

Figure 1. Automated segmentation and morphology assessment of the hip joint from MRI scans of the Generation R cohort.

In addition, together with orthopedic surgeon Jaap Tolk, we are assessing hip morphology in young adults with hip dysplasia before and after corrective surgery using our automated morphology assessment tools. In another project utilizing Generation R MRI scans, together with researchers from Tübingen, Germany, we have developed methods to automatically segment and extract biomarkers from upper leg muscles in adolescents. We are also actively involved in a project assessing knee shape in adolescents using the Generation R data. This research project is a collaboration with the department of General Practice (Rosemarijn van Paassen, Marienke van Middelkoop).

Image analysis of advanced imaging techniques We are collaborating closely with HIT CoC (Gyula Kotek) and ADMIRE group (Edwin Oei, Rianne van der Heijden) to develop automated tools for detecting abnormal regions in PET-MRI. This project involves automated segmentation of various structures on MRI (Figure 3), synthetic PET lesion generation to expand the training dataset, and the development of machine learning and deep learningbased detection algorithms.

Figure 3. Automatically segmented spinal MRI scan and PET signal in the spinal canal.

scientific report 2025 | MUSCULOSKELETAL IMAGE ANALYSIS

Miscellaneous

Additional Personnel

Quantitative imaging is an essential tool for evaluating musculoskeletal conditions. However, the diversity in imaging modalities, sites, and vendors makes comparison and pooling of data impossible without standardized, open-source post-processing pipelines. I am a member of Open and Reproducible Musculoskeletal Imaging Research (ORMIR) community (www.ormir.org), which is dedicated to advancing open and reproducible musculoskeletal imaging research. As part of this effort, I actively contribute to the ORMIR-MIDS initiative, which aims to establish an open standard for curating and sharing musculoskeletal imaging data.

Yijie Fang – MD, guest postdoctoral researcher from China

Expectations & Directions

Delft, and Mathematical Sciences, Utrecht UniversityFloris Rupert – MSc student Applied Physics, TU Delft

In the coming years, we aim to further develop AI-based techniques for analyzing PET-MRI and photon-counting CT images, with a focus on musculoskeletal applications. With automated AI-based image analysis tools, we can extract a variety of information from population studies at Erasmus MC. This enables investigation of biomarkers and disease processes to enhance our understanding of musculoskeletal diseases. The new Motion Biomechanics and Imaging Lab will open many new research directions and further strengthen our collaboration with TU Delft.

Funding Hirvasniemi, Jukka, Edwin Oei, and Rianne van der Heijden TU Delft-Erasmus MC Convergence Flagship: ‘Healthy Joints’. Main applicant: S.M.A. Bierma-Zeinstra (General Practice), J. Harlaar (TU Delft). 2023-2027 Hirvasniemi, Jukka and Edwin Oei European Research Council (ERC) Advanced grant: ‘Biomechanical precision diagnostics in osteoarthritis Modelling trajectories and mechanisms of childhood hip dysplasia’. Main applicant: S.M.A. Bierma-Zeinstra (General Practice/Orthopedics). 2023-2027

Rosemarijn van Paassen – PhD student Teemu Nurmirinta – visiting PhD student from UEF, Finland Netanja Harlianto – MSc student Medicine, University of Utrecht Lucas Bronder – MSc student NIHES Nick Hanenberg – MSc student Biomedical Engineering, TU Delft Martijn Brouwer – MSc student Applied Physics, TU

Wesley de Reus – MSc student Technical Medicine, TU Delft Timothy Seijerling – MSc student Technical Medicine, TU Delft Jauke Engels – MSc student Technical Medicine, TU Delft Febricetta Sarwono – MSc student Biomedical Engineering, TU Delft Khalil Akhundov – MSc student Biomedical Engineering, TU Delft Bo Verduijn – MSc student Biomedical Technology and Physics, VU Amsterdam Keira Zieboll – MSc student Artificial Intelligence, VU Amsterdam Yutong Wu – MSc student Biomedical Engineering, TU Delft Mark Planken – MSc student Mechanical Engineering, TU Delft

PhD Student

Invited Lectures

Mirthe Kamphuis, MSc

Jukka Hirvasniemi. ‘AI-based lesion detection on PET/MRI’. PET/MRI-CoC Symposium, Rotterdam, the Netherlands. February 2025.

Advisors Edwin Oei, Stefan Klein, Jukka Hirvasniemi & Jos Runhaar

Jukka Hirvasniemi. ‘MSK image analysis’. EMC-UWMadison Symposium, Rotterdam, the Netherlands. March 2025.

The Impact of Hip Dysplasia on the Young Adult Hip Joint Hip dysplasia significantly influences joint health and resilience. This study examines the role of hip dysplasia in combination with osteoarthritis (OA) risk factors, including genetic predisposition, obesity, and physical activity levels. Through morphological biomarkers and radiomics, the research aims to uncover early markers of joint deterioration.

Mirthe Kamphuis. ‘Automatic 3D analysis of hip morphology in a young adult population’. Sophia Research Day, Rotterdam, the Netherlands. April 2025.

Martijn PA Starmans, PhD assistant professor 88

ARTIFICIAL INTELLIGENCE FOR INTEGRATED DIAGNOSTICS Top Publications 2025

JOINT APPOINTMENT IN PATHOLOGY Dr. ir. Martijn Starmans is Assistant Professor with a joint appointment at the Department of Pathology & Bioinformatics and is heading the AI for Integrated Diagnostics (AIID) research line. His vision is that we can and should learn more from previous studies, which he pursues by working on meta-level methods across clinical applications. His current research interests include radiomics, pathomics, multimodal machine learning, meta-learning, and trustworthy AI, with a focus on application in oncology. In 2022, Martijn obtained his PhD degree “cum laude” with his thesis on Streamlined Quantitative Biomarker Development at the Erasmus MC. He is recipient of an NGF AINed Personal Fellowship 2023, and co-principal investigator of the funded Sarcoma AI (SAI) consortium and the Liver AI (LAI) consortium. He has a strong affinity with research infrastructure, being a work-package lead in the Horizon EuCanImage, EOSC4Cancer, and AFRICAI-RI projects. Additionally, he is co-founder and co-lead of the MICCAI Open Data initiative to promote and facilitate open science. Martijn actively collaborates with the BCN-AIM group of prof. Lekadir at the University of Barcelona, with whom he created the FUTURE-AI guideline for trustworthy AI.

Lekadir K, AF Frangi, AR Porras, B Glocker, C Cintas, CP Langlotz, E Weicken, FW Asselbergs, F Prior, GS Collins, G Kaissis, G Tsakou, I Buvat, J Kalpathy Cramer, J Mongan, JA Schnabel, K Kushibar, K Riklund, K Marias, LM Amugongo, LA Fromont, L Maier Hein, L Cerdá Alberich, L Martí Bonmatí, MJ Cardoso, M Bobowicz, M Shabani, M Tsiknakis, MA Zuluaga, M C Fritzsche, M Camacho, MG Linguraru, M Wenzel, M de Bruijne, MG Tolsgaard, M Goisauf, M Cano Abadía, N Papanikolaou, N Lazrak, O Pujol, R Osuala, S Napel, S Colantonio, S Joshi, S Klein, S Aussó, WA Rogers, Z Salahuddin, MP Starmans. FUTURE AI: International consensus guideline for trustworthy and deployable artificial intelligence in healthcare. BMJ 2025; 388: e081554. Spaanderman DJ, M Marzetti, X Wan, AF Scarsbrook, P Robinson, EH Oei, JJ Visser, R Hemke, K van Langevelde, DF Hanff, GJ van Leenders, C Verhoef, DJ Grünhagen, WJ Niessen, S Klein, MP Starmans. AI in radiological imaging of soft tissue and bone tumours: A systematic review evaluating against CLAIM and FUTURE AI guidelines. EBioMedicine 2025; 114:105642. Kondylakis H, R Osuala, X Puig Bosch, N Lazrak, O Diaz, K Kushibar, I Chouvarda, S Charalambous, MP Starmans, S Colantonio, N Tachos, S Joshi, HC Woodruff, Z Salahuddin, G Tsakou, S Aussó, L Cerdá Alberich, N Papanikolaou, P Lambin, K Marias, M Tsiknakis, DI Fotiadis, L Martí Bonmatí, K Lekadir. A review of methods for trustworthy AI in medical imaging: The FUTURE AI Guidelines. IEEE Journal of Biomedical and Health Informatics 2025; 1–18.

m.starmans@erasmusmc.nl

Context

rtificial Intelligence (AI) has made substantial advances in predictive models for precision medicine. Within this field, medical imaging has gained an increasingly important role, with two of the most important domains being radiology (“radiomics”) and pathology (“pathomics”). While these often have similar goals and contain complementary information, these research fields are largely separated and rarely combined. Additionally, AI models are typically developed from scratch, considering only disease-specific datasets: knowledge on what worked well in previous AI applications is largely ignored. In the AI for Integrated Diagnostics (AIID) we develop and evaluate novel multimodal machine learning methods to create quantitative biomarkers, with a focus on medical imaging and application in oncology. By developing pancancer methods on a meta-level across diseases, we facilitate generalization of our methods to other clinical domains. We strongly collaborate with clinicians and external partners in inter-disciplinary consortia to ensure trustworthy biomarkers that can truly aid clinicians.

Research Projects: Objectives & Achievements

trustworthy and deployable AI In the BMJ, an effort of 117 Interdisciplinary experts which Dr. Starmans co-led. In collaboration with Dr. Visser (page 224), Erik Kemper published a position paper on the role and need for early health technology assessment (eHTA) for value-based radiology AI. He Is finishing an eHTA method for radiology AI utilizing the FUTURE-AI guideline, which he presented at the ECR 2025 conference.

Novel multimodal machine learning methods Together with Dr. Klein (page 64), we established a successful collaboration between the department of Radiology & Nuclear Medicine and Pathology, uniting BIGR and the PHANTOM group in the AIID research line. Our mission is to enable research on biomarkers integrating various sources of data using artificial intelligence, with a focus on radiological and pathological imaging. To this end, Fariba Tohidinezhad has developed a crossattention multimodal machine learning method which is being applied for predicting targeted therapy response in gastrointestional stromal tumors. To automate design choices, Natalia Oviedo Acosta has developed an algorithm for automated method design in cancer Image classification, which has been presented at the MSB Emerge workshop at the MICCAI 2025 conference In SouthKorea and published in the proceedings. She Is currently working on a method to Improve both the performance and efficiency of AI model training using knowledge from previous applications in medical Imaging, In collaboration with the AutoML group of the University of Freiburg. Rick Bortsov started their PostDoc on trustworthy AI In multimodal machine learning for radiopathomics. They are currently working on raising awareness and an evaluation framework for shortcut learning.

Quantitative imaging biomarkers in oncology In collaboration with Dr. Klein, our sarcoma research line has expanded and made substantial advancements. Douwe Spaanderman, Xinyi Wan, and Matthew Marzetti performed a systematic review on AI for bone and soft tissue tumors with a focus on trustworthiness, utilizing the FUTURE-AI guideline, to create a roadmap for this research line and the translation of these methods to clinical practice. Douwe Spaanderman developed a method for predicting failure of active surveillance In desmoid-type fibromatosis on MRI, which was presented at the ESMO 2025 conference. In parallel, he developed a benchmark for extracting structured data from narrative reports using large language models (LLMs). Xinyi Wan is currently working on a method for predicting treatment response to neoadjuvant chemotherapy in osteosarcoma on MRI. Additionally, she is conducting a research visit at La Fe, Valencia, to develop a method for diagnosing all bone tumors on MRI. Matthew Marzetti presented his work on using IVIM and ADC parameters to distinguish

Figure 1. In the artificial intelligence for integrated diagnostics (AIID) research line, we develop novel multimodal machine learning methods to integrate primary radiology and pathology data to develop trustworthy biomarkers, focused on oncology.

Trustworthy and value-based AI for clinical practice Besides biomarker development, we focus on trustworthiness to enable transition of AI models to clinical practice, which includes our involvement in the Trustworthy AI for MRI ICAI lab of Dr. Poot (page 72) through the work of Xinyi Wan. A major highlight in 2025 was publication of the FUTURE-AI guideline for

malignant sarcoma on MRI at ISMRM 2025, which work Is he currently extending. Karthik Prathaban presented his work on combining knowledge graphs with LLMs at the ELAMI workshop at the MICCAI 2025 conference, which is published in the proceedings. Besides extending the LLM-work, he Is also currently working on a method for automated phenotyping and grading of soft tissue tumors on whole slide imaging using AI.

scientific report 2025 | ARTIFICIAL INTELLIGENCE FOR INTEGRATED DIAGNOSTICS

In collaboration with Dr. Thomeer (page 178) and Dr. Klein, Dr. Starmans is leading the Liver Artificial Intelligence (LAI), which aims to fast-forward the development, validation, and implementation of machine learning methods that could support MRI-based diagnosis of liver lesions. A dedicated website was launched: https://laiconsortium.org/. Ruben Niemantsverdriet and Frederik Hartman presented our consortium at the ESGAR and EASL 2025 conferences. They are currently working on the collection and curation of a benchmark liver cancer MRI dataset, the LAI research Infrastructure, and an eHTA for liver cancer MRI employing the method from Erik Kemper.

Lastly, dr. Starmans continued his work as co-lead of the MICCAI Open Data initiative, which he co-founded in 2024 to facilitate and promote the sharing of medical imaging datasets. In this context, he was part of the MICCAI 2025 organization. As part of the track, they organized a main event session, a special issue Including publication of 13 new datasets. For their work, the Initiative won a MICCAI Health Equity Award.

Expectations & Directions We will further expand both the fundamental method development side, as well as the applied research on developing disease-specific biomarkers in our current and “new” clinical applications. As we envision further integration of additional omics data besides radiology and pathology, ultimately, we are aiming for an integrated diagnostics approach, using AI to combine all relevant data in the most optimal way. Hence, we aim for additional experts to join our multi-disciplinary AIID research line.

Together with the department of Surgical Oncology and Dr. Klein, Zhen Qian developed a method for automated growth pattern segmentation on whole slide imaging pathology of colorectal liver metastases, which he presented at CLINICCAI at the MICCAI 2025 conference. He Is currently expanding this approach to find additional histopathological biomarkers for survival prediction. In collaboration with the Erasmus MC Center for Pain, Eline van Lange developed an AI method to diagnose complex regional pain syndrome (CRPS) from video thermography. She Is currently working on evaluating the model In CRPS treatment response prediction. We established a collaboration successful collaboration with Dr. Zuiverloon of the department of Urology, Dr. Akram of the department of Pathology & Bioinformatics, and Dr. Khalili of the Radboud UMC, on artificial intelligence-driven molecular subtype classification of urothelial cancer from histopathological images. In this project, Tongjie Wan started his PostDoc, and Sepideh Bahri Hampa her PhD.

Infrastructure for cancer research To perform our research, infrastructure for e.g., data collection, storage, annotation, harmonization, integration, and analysis is essential. We are also involved in research infrastructure on a larger scale in close collaboration with Dr. Klein, Dr. Bron (page 76), and Hakim Achterberg. In the H2020 EuCanImage project, we finalized the biobanking of imaging and clinical data of 25.000 patients across Europe. In the H2021 EOSC4 Cancer project, we finished a framework for harmonization and integration of various data types (e.g., clinical, radiology, pathology, genomics) for cancer research. In 2025, the EU Horizon AFRICAI RI project started with the goal to build a pan African research infrastructure for medical Image analysis with AI, where we lead the Infrastructure work package. In this project, Mahlet Birhanu started as PhD student, and Aisha Goedhart as research software engineer.

Funding Lekadir, Karim, Aad van der Lugt, Wiro Niessen, Stefan Klein, Daniel Bos, and consortium partners EU Horizon2020: ‘EUCanImage - A European Cancer Image Platform Linked to Biological and Health Data for Next-Generation Artificial Intelligence and Precision Medicine in Oncology’. 2020-2025 Klein, Stefan, Jan-Jaap Visser, Dirk Grunhagen, Kees Verhoef, Stefan Sleijfer, Wiro Niessen, Arno van Leenders, and Martijn Starmans Hanarth Fonds: ‘Automatic grading and phenotyping of soft-tissue tumors through machine learning to guide personalized cancer treatment’. 2021-2025 Valencia, Alfonso, Stefan Klein, Aad van der Lugt, Wiro Niessen, Marcel Koek, and consortium partners EU HORIZON INFRA-2021-EOSC-01: ‘EOSC4CANCER: A European-wide foundation to accelerate Data-driven Cancer Research’. 2022-2025 Marti-Bonmati, Luis, Esther Bron, Martijn Starmans, Marcel Koek, Jan-Jaap Visser, Wiro Niessen, Stefan Klein, and consortium partners EU HORIZON DIGITAL-2022-CLOUDAI-02: ‘EUCAIM: European Federation for Cancer Images’. 2023-2027 Marzetti, Matthew NIHR Research Design Service for Yorkshire and Humber Public Involvement Grant: ‘Novel Applications for Sarcoma Assessment (NASA)’. 2023-2026 Visser, Jan-Jaap, Martijn Starmans, Ken Redekop, and Frans Vos Health-Holland LSH-TKI PPP: ‘An artificial intelligence (AI)-based model for detection of incidental pulmonary embolism in chest CTs’. 2023-2027 Niessen, Wiro, Aad van der Lugt, Juan Hernandez-Tamames, Stefan Klein, Edwin Oei, Dirk Poot, Marion Smits, JanJaap Visser, and consortium partners NWO Min EZK and GE

healthcare: ‘ROBUST consortium: Trustworthy AI for MRI ICAI lab’. 2022-2027 Klein, Stefan, Marcel Koek, Hakim Achterberg, Adriaan Versteeg, Martijn Starmans, and consortium partners ISIDORe JRA PROGRAMME: ‘PATH2XNAT: Covid19 Pathomics meets XNAT’. 2023-2025 Klein, Stefan, Maarten Thomeer and Martijn Starmans NWO Open Technology Programme: ‘The Liver Artificial Intelligence (LAI) consortium: a benchmark dataset and optimized machine learning methods for MRI-based diagnosis of solid appearing liver lesions’. 2024-2028 Starmans, Martijn NGF AiNed Fellowship Grant: ‘Radiology and pathology join forces through Artificial Intelligence for Integrated Diagnostics (AIID)’. 2024-2029 Visser, Jan-Jaap, Ilva van Houwelingen, Martijn Starmans, and Stefan Klein EU Horizon DIGITAL-2024-CLOUD-DATA-06: ‘SHAIPED: Supporting Health Data Access Bodies to establish AI pathways enabling Deployment of AI as medical device tools’. 2024-2027

Faber Pas – Internship student Danchen Ge – Internship student Qiang Zhao – Internship student Jialin Song – Internship student Nicolas Jager Gallego - Internship student Carlo Peron - Internship student Amal Lahchim - Internship student Junkang Wei - Internship student Wanyi Lin - Internship student Timothy Seijerling - Internship student

Invited Lectures Frederik Hartmann, Ruben Niemantsverdriet, Stefan Klein, Maarten Thomeer and Martijn Startmans. ‘Advances of artificial intelligence as a diagnostic aid in abdominal radiology: The liver artificial intelligence (LAI) consortium’. Dutch Liver Retreat, Lunteren, the Netherlands. Feb 2025. Martijn Starmans. ‘The FUTURE-AI guideline: How to design you’re AI tool in a trustworthy way towards clinical value and implementation’. AI Symposium - Applied AI in Clinical Research: From Concept to Clinical Impact, Rotterdam, the Netherlands. July 2025.

Additional Personnel Stavros Makrodimitris - PostDoc Tongjie Wang - PostDoc Douwe Spaanderman - PhD student Erik Kemper - PhD student Ruben Niemantsverdriet - PhD student Frederik Hartmann - PhD student Eline van Lange - PhD student Sepideh Bahri Hampa - PhD student Mahlet Birhanu – Research Software Engineer Aisha Goedhart – Research Software Engineer

scientific report 2025 | ARTIFICIAL INTELLIGENCE FOR INTEGRATED DIAGNOSTICS

Post-docs

Rick Bortsov, PhD Trustworthy AI for medical image analysis in radiology and pathology

Developing artificial intelligence that can be safely and confidently used in clinical radiology and pathology workflows requires more than strong predictive performance. AI systems must be reliable, transparent, and robust across different scanners, hospitals, and patient populations. This project contributes to the broader effort of building trustworthy multimodal machine learning models. The goal is to ensure that these models make decisions based on meaningful disease‑related patterns rather than incidental characteristics of the data, ultimately supporting more dependable diagnostic tools in oncology. A current focus within this work is understanding and mitigating “shortcut learning,” a phenomenon where AI models latch onto unintended cues-such as imaging artifacts or dataset imbalances-instead of true biological signals. By developing methods to detect these shortcuts and designing strategies to prevent them, the project aims to strengthen the reliability of radiopathomics models and support their translation into clinical workflows. This research aligns with emerging international guidelines for trustworthy AI and contributes to building a foundation where advanced machine learning can be integrated into clinical practice with confidence.

Fariba Tohidinezhad, PhD Multimodal machine learning for integrated radiomics and pathomics

Integrating the combined potential of radiomics and pathomics through multimodal machine learning offers a transformative approach to understanding cancer biology and improving prediction of patient outcomes. Radiomics extracts quantitative features from medical imaging, while pathomics derives high-dimensional data from histopathological slides. Individually, these modalities provide valuable insights into tumor characteristics, but their integration can capture complementary information critical for precise diagnosis and prognosis. Multimodal learning enables the fusion of these data types by leveraging advanced machine learning techniques, such as attention mechanisms, to model complex relationships between radiology and pathology. This project focuses on developing and validating robust multimodal frameworks using sarcoma as a use case, aiming to enhance the predictive performance of diagnostic and prognostic models. The approach seeks to bridge the gap between radiology and pathology, addressing the fragmented nature of current data analyses. Ultimately, we aim to set a foundation for scalable applications of multimodal learning across diverse clinical scenarios, advancing precision oncology.

PhD Students Karthik Prathaban, MEng

Xinyi Wan, MSc

Advisors Martijn Starmans, Farhan Akram & Stefan Klein

Advisors Stefan Klein, Martijn Starmans, Edwin Oei & Jacob Visser

Advancing Digital Pathology for Soft Tissue Tumors (STTs)

Develop Trustworthy AI methods for Improved Diagnosis of Bone and Soft-tissue lesions on MRI

Digital pathology studies for soft tissue tumors (STTs) are limited and often constrained by small datasets and unreliable computational frameworks. This project aims to develop comprehensive, pre-operative pathologybased models, leveraging AI, traditional computer vision methods, or both, to enhance diagnostic accuracy and support treatment planning for better clinical outcomes.

MRI plays a pivotal role in the diagnosis of bone and softtissue lesions. However, the wide variety of lesions share clinical manifestations. To address this, we propose the implementation of AI-based imaging analysis algorithms to aid healthcare. Additionally, we will prioritize the trustworthiness of AI models to ensure interpretable results.

Matthew Marzetti, MSc

Mahlet Birhanu, MSc

Advisors

David Buckley, Andrew Scarsbrook, Martijn Starmans & Stefan Klein

Martijn Starmans, Hakim Achterberg, Esther Bron & Stefan Klein

AI-assisted diagnosis of soft tissue tumors using radiomics and quantitative MRI

AFRICAI RI: A Pan African Research Infrastructure for Collaborative Biomedical Imaging and Artificial Intelligence

This project aims to develop AI models to help differentiate benign and malignant soft tissue masses using both routinely acquired clinical MR images and quantitative MR images. If successful, this research has the potential to reduce the number of patients requiring soft tissue biopsies for the diagnosis of soft tissue tumors.

Towards Inclusive AI, there Is currently a lack In African high‑quality imaging dataset, standardized workflows, and the infrastructure needed to develop medical AI. This project addresses these gaps by building a pan-African research Infrastructure, focusing on federated learning and trustworthy AI, to empower and upscale African medical Imaging research.

Natalia Oviedo Acosta, MSc

Zhen Qian, MD

Advisors Martijn Starmans & Stefan Klein

Advisors Dirk Grunhagen, Kees Verhoef, Martijn Starmans & Stefan Klein

Translating Pan-Cancer Model Knowledge Across Diseases with Automated Machine Leaning and Meta-Learning

Artificial intelligence in histopathological whole-slide images of colorectal liver metastases

AI models for medical imaging are often developed from scratch with limited data, making them costly and prone to overfitting. To address this, we propose pan-cancer metalearning to translate knowledge across clinical domains, integrating AutoML to tackle challenges in architecture selection, imaging adaptation, and computational cost, creating generalizable solutions.

To advance diagnosis, survival prediction, and personalized therapy response, this project aims to develop AI models based on histopathology to automate and enhance biomarkers related to histopathological growth pattern of colorectal liver metastases.

scientific report 2025 | ARTIFICIAL INTELLIGENCE FOR INTEGRATED DIAGNOSTICS

Danilo Andrade De Jesus, PhD assistant professor

Luisa Sanchez Brea, PhD assistant professor 96

EYE IMAGE ANALYSIS GROUP ROTTERDAM

Top Publications 2025

JOINT APPOINTMENT IN OPHTHALMOLOGY AND EYE HOSPITAL ROTTERDAM

Katarya K, VA de Vries, J Barbosa-Breda, LS Brea, DA de Jesus, WD Ramdas. Prognostic factors of optical coherence tomography angiography in intermediate and advanced primary open-angle glaucoma: A systematic review and meta-analysis. Acta Ophthalmologica 2025; 10.1111.

The Eye Image Analysis Group Rotterdam (EyeR) is a collaboration between the departments of Radiology & Nuclear Medicine and Ophthalmology of Erasmus MC, together with the Rotterdam Eye Hospital. Led by Principal Investigators Luisa Sanchez Brea and Danilo Andrade De Jesus, the group focuses on applying artificial intelligence and advanced image processing techniques to ophthalmology. The overall ambition of the EyeR group is to improve the diagnosis and monitoring of retinal diseases by developing innovative methodologies that make use of multi-scale and multi-modal imaging. By integrating advanced imaging data with novel analytical approaches, the group aims to extract deeper insights into retinal microstructures and disease mechanisms. Key research interests of the group include motion correction and compensation for imaging artifacts in Optical Coherence Tomography data, the development of end-to-end models for the staging of retinopathy of prematurity using clinical data, and the application of cutting-edge technologies such as Adaptive Optics to advance the study of rare conditions like inherited retinal dystrophies, rhegmatogenous retinal detachment recovery, and glaucoma.

M Pazos, A Tatham, D Andrade De Jesus, L Sanchez Brea, S Klein, D Wright, ... A Viswanathan. Realworld data landscape for glaucoma in Europe: a questionnaire-based analysis of resources among European Glaucoma Society members. BMJ Open Ophthalmology 2026; 104: e491-e502. Kashani AH, TT Berendschot, S Bonnin, L Sobrin, DA De Jesus, L Sanchez Brea, ... FC van der Heide. Retinal optical coherence tomography angiography imaging in population studies for study of microvascular dysfunction in Alzheimer’s disease and related dementias. Alzheimer’s & Dementia 2025; 21, e70252. Wooning S, PAT Heutinck, K Liman, S Hennekam, M van Haute, F van den Broeck, B Leroy, DM Sampson, D Roshandel, FK Chen, DM Pelt, LI van den Born, VJM Verhoeven, CCW Klaver, AA Thiadens, M Durand, N Chateau, T van Walsum, D Andrade De Jesus, L Sanchez Brea. Automated Cone Photoreceptors Detection in Adaptive Optics Flood-Illumination Ophthalmoscopy. Ophthalmology Science 2024; 5: 100675.

m.sanchezbrea@erasmusmc.nl d.andradedejesus@erasmusmc.nl

Context

he human eye serves a dual purpose: it provides a visual reference to our surroundings while also acting as a window into our overall health. Its unique optical structure enables direct, non-invasive observation of blood vessels, nerves, and tissues, offering insights into ophthalmic, systemic, and neurological conditions. This makes the eye interesting for diagnosing and monitoring diseases such as diabetes, hypertension, and neurodegenerative disorders. High-resolution imaging technologies like optical coherence tomography (OCT), OCT angiography (OCTA), and adaptive optics (AO), are becoming increasingly common in both research and clinical settings. Subtle changes in light reflectivity, attenuation, and scattering can serve as early indicators of disease and help monitor its progression over time. To process and analyze such data in an objective, repeatable, and efficient manner, artificial intelligence (AI) and advanced image processing play a major role. Automated analysis of retinal images has proven effective in detecting diseases, grading their severity, and predicting progression. Moreover, retinal image analysis assisted by AI holds the potential to uncover novel mechanisms underlying various conditions, driving advancements in both diagnostics and treatment strategies.

Research Projects: Objectives & Achievements

cells in the retinal pigment epithelium. AO also provides a more detailed view of disease markers, including drusen, microaneurysms, or atrophic regions in the retina.

Analysis of OCTA images of the retina Optical Coherence Tomography Angiography (OCTA) is a non-invasive imaging modality that enables the observation of the blood flow within the retina. The retina plays a vital role in vision, and many systemic and ophthalmic diseases, such as diabetes, glaucoma, or optic neuritis, are associated with damage to the retinal vasculature. Therefore, studying the information in the OCTA images can aid in monitoring these diseases, identifying optimal treatment windows, and improving our understanding of disease mechanisms, such as in detecting preclinical changes. However, OCTA images are highly detailed, and visual assessment alone is not sufficient to detect subtle variations, which are often the most clinically relevant for timely intervention. Thus, there is a need to objectively “summarize” the information contained in these images. With the help of artificial intelligence, automated methods can be developed to quantify specific biomarkers, such as microvascular density, across different regions of the retina (e.g. around the foveal avascular zone or the optic disc) and within specific sectors of the image (superior, inferior, temporal, nasal). These biomarkers can then be used to model the disease progression and to detect deviations from what is expected in a normal, healthy retina.

AO has been widely applied in age-related macular degeneration, hypertension, and diabetes, and there is a growing interest in its use for glaucoma and in inherited retinal dystrophies. Although the technology is primarily used in research settings, recent studies have begun to incorporate AO imaging in clinical practice, particularly for monitoring treatment effects, such as in gene therapy trials. AO offers exciting opportunities for retinal imaging, with significant potential for earlier and more accurate detection of pathological changes. This, in turn, can support personalized treatment strategies and improve patient follow-up. However, to make use of these advantages, automated analysis tools are needed to quantify specific image features (e.g., cell counts or vessel wall thickness) ensuring that subtle changes are reliably detected.

Analysis of specular microscopy images of the corneal endothelium The cornea is the transparent, dome-shaped outer layer of the eye that helps focus light as it enters the eye, contributing to the eye’s overall focusing power. Its innermost layer, the corneal endothelium, has a regulatory function to maintain transparency, ensuring proper vision. Unlike other cells, corneal endothelial cells do not significantly regenerate and, if many of them are damaged, it will lead to consequences for our vision. Some cells are lost naturally with age, but there are also several factors such as diseases, direct injuries, or incorrect wear

Figure 1. OCTA image of the superficial vascular plexus of the retina. (A) Original image, (B) Result of a model trained for automatic segmentation of the vasculature.

Adaptive Optics imaging for detailed analysis of retinal structures The retina is a highly complex tissue composed of multiple layers, each consisting of distinct cellular and structural components. Adaptive Optics (AO) is an advanced technique that enables visualization of retinal structures that cannot be resolved using conventional imaging methods, such as cone photoreceptors, vessel walls, or the

Figure 2. Adaptive Optics image of a healthy human retina. Adaptive Optics enables visualization of individual cone photoreceptors.

scientific report 2025 | EYE IMAGE ANALYSIS GROUP ROTTERDAM

of contact lenses, which can cause alterations and cell loss. Furthermore, common surgeries, such as in cataract or glaucoma, are also known to affect corneal endothelial cells. Therefore, it is important to monitor them to ensure timely clinical intervention, preventing permanent damage. Specular microscopy is an advanced imaging technique that provides a high-resolution visualization of the corneal endothelium. While these images are routinely acquired in clinical practice, and the acquisition device provides itself some basic measurement tools, these tools often lack accuracy and may show discrepancies when we compared with manual counts performed by human experts. Also, there are more structures that can be measured in the images, such as corneal guttae, edema, or lesions. These are especially important for tracking the progression of certain diseases, such as Fuchs’ Endothelial Corneal Dystrophy.

Expectations & Directions EyeR’s PIs have established a global network of collaborators, including clinical centers, AI experts, and industry partners. In the short term, the group is working closely with these collaborators to expand its current research lines. Key objectives include multi-center studies aimed at collecting normative data using Adaptive Optics and OCTA, as well as large-scale research initiatives, such as collaboration with University of Porto (Portugal) and Medical University of Bialystok (Poland). In addition, EyeR is involved in population-based research of OCTA and AO data within the Maastricht Study, and the use of real-world data through the collaboration with the Rotterdam Eye Hospital. Furthermore, EyeR is pioneering Adaptive Optics research in the Netherlands, leading collaborative efforts with UMC Maastricht and Radboud University in this domain. Looking ahead, EyeR aims to further expand the scope of ophthalmic imaging beyond traditional ophthalmology. In particular, the group is interested in exploring the relationship between retinal imaging biomarkers and cardiovascular health, opening new avenues for interdisciplinary research and clinical applications.

Funding Andrade De Jesus, Danilo, Hans Lemij, Pieter Bonnemaijer Stichting Wetenschappelijk Onderzoek het Oogziekenhuis: ‘Added value of Adaptive Optics Ophthalmoscopy in the management of glaucoma: a pilot study’. 2024-2026 Van Romunde, Saskia, and Danilo Andrade De Jesus Stichting Wetenschappelijk Onderzoek het Oogziekenhuis: ‘Adaptive optics in patiënten na netvliesloslating’. 20242025 Derks, Lizanne, Jarinda Poppe, Angela Arends -Tjiam, Rob Taal, Danilo Andrade De Jesus, Luisa Sanchez Brea Wishdom Foundation: ‘Voorspellen van ernstige prematurenretinopathie met behulp van kunstmatige intelligentie: uitbreiding en klinische validatie van een werkend model’. 2023-2026 Van Walsum, Theo, Caroline Klaver, Nicolas Chateau, Danilo Andrade De Jesus, Luisa Sanchez Brea Health Holland TKI Call: ‘AO-Vision - Adaptive Optics imaging: A guiding star to save vision’. 2022-2026

Figure 3. Specular microscopy image of corneal endothelium of a patient. The prediction of an automatic model for guttata detection and the manual annotations of a human grader are overlapped on the image.

Invited Lectures Danilo Andrade De Jesus, “Introduction to Ophthalmic Imaging: Principles & Applications”, Kinderhersenlab, Erasmus MC Sophia Kinderziekenhuis, Rotterdam, the Netherlands. Feb 2025. Danilo Andrade De Jesus. “AI analytics for AO imaging biomarkers”, Association for Research in Vision and Ophthalmology Annual Meeting, Salt Lake City, USA. May 2025. Danilo Andrade De Jesus, “How much knowledge of AI do optometrists really require?” 22nd Edition of the International Congress of Optometry and Vision Science , University of Minho, Braga, Portugal. May 2025.

Additional Personnel Davide Pasqualini (Technical Medicine Internship) Eva Horlings (Technical Medicine Internship) Dimitrios Marinos (MSc internship) Bastian Tostivint (MSc internship) Aniek Sips (MSc thesis)

PhD Student Adriana Falcão Neves, MSc Advisors Luisa Sánchez Brea, Danilo Andrade De Jesus, Theo van Walsum & João Barbosa Breda

A novel OCTA framework for ophtalmic and systemic diseases assessment Enhance the usability of OCTA data, by overcoming imaging artifacts and creating automatic tools for quantifying retinal biomarkers. Hence enabling earlier diagnosis, improved disease monitoring, prognosis and better patient outcomes.

100

scientific report 2025 | EYE IMAGE ANALYSIS GROUP ROTTERDAM

101

102

Focus Area

MOLECULAR IMAGING & THERAPY The aim is to study molecular and cellular events in a non-invasive manner and to develop new treatment modalities for cancer. The research focuses on the development of contrast agents, reporter genes, radiopharmaceuticals and multimodality agents for MRI, optical and/ or radionuclide imaging and therapy, as well as their functioning within the cell and/or whole organism, their preclinical validation, and their clinical translation, to ultimately improve the cure rate and quality of life of patients.

103

Julie Nonnekens, PhD associate professor 104

RADIOBIOLOGY OF RADIONUCLIDE THERAPY

JOINT APPOINTMENT MOLECULAR GENETICS

Top Publications 2025

Julie Nonnekens received her MSc in Biotechnology at Wageningen University in 2009 and completed her PhD in cancer biology, focusing on DNA repair mechanisms, at the University of Toulouse (France) in 2013. Following, she was a postdoc at the Hubrecht Institute working on ribosome biogenesis in cancer and longevity. Since 2014, Julie has been at Erasmus MC, holding a dual appointment in Radiology & Nuclear Medicine and Molecular Genetics. Her research integrates DNA damage repair and nuclear medicine to explore the radiobiology of targeted radionuclide therapy, aiming to optimize treatment strategies.

Engbers PA, TG Reuvers, JM Heredia-Genestar, J Chang, NS Verkaik, M Sabatella, J Nonnekens. Radiosensitization of NET cells by HSP90 inhibitor ganetespib is mediated through pleiotropic stress responses. EJNMMI Res. 2025; 15:146.

Julie has received multiple recognitions, including young investigator awards, and serves as principal investigator on prestigious grants, such as the ERC starting grant. She chairs the Netherlands Society of Radiobiology and co-founded the European working group on Radiobiology of Molecular Radionuclide Therapy. Her work fosters interdisciplinary collaboration to advance cancer treatment through innovative approaches in radiobiology and nuclear medicine.

Mendes F, SY Terry, D Spiegelberg, B Cornelissen, J Bolcaen, J Nonnekens. European Working group of Radiobiology of Molecular Radionuclide Therapy. Recommendations for reporting preclinical radiobiological studies in targeted radionuclide therapy. Eur J Nucl Med Mol Imaging 2025; 52:30663070.

Tamborino G, P Engbers, TH de Wolf, TG Reuvers, R Verhagen, M Konijnenberg, J Nonnekens. Establishing in vitro dosimetric models and dose–effect relationships for 177Lu-DOTATATE in neuroendocrine tumors. Journal of Nuclear Medicine 2025; 66: 1291-1298.

j.nonnekens@erasmusmc.nl

Context

argeted radionuclide therapies (TRT) are transforming the treatment of metastasized cancers by delivering radiolabeled compounds directly to tumor cells via specific receptors. Once bound, the radionuclides emit ionizing radiation, causing DNA damage leading to cancer cell death. Despite growing clinical application, challenges remain, as some patients experience over-treatment (toxicity) or under-treatment (insufficient tumor regression), emphasizing the need for optimization. Advancing our understanding of the radiobiology of TRT – the biological effects of ionizing radiation – and integrating this knowledge into improved dosimetric models are critical steps toward enhancing therapeutic outcomes. By providing evidence to guide treatment methods and regimens, these efforts aim to refine TRT approaches, reducing toxicity and improving efficacy. Our overarching goal is to bridge radiobiological research with clinical practice, driving the evolution of TRT from a primarily palliative to a potentially curative treatment. For more information about our research, visit www.nonnekenslab.com.

105

Research Projects: Objectives & Achievements Cellular effects of TRT in tumor cells Our research focuses on radiolabeled compounds, particularly those using the β-emitter lutetium-177, such as somatostatin analogue DOTA-(Tyr3)octreotate ((177Lu) Lu-DOTA-TATE) for neuroendocrine tumors (NET) and prostate-specific membrane antigen (PSMA) compounds ((177Lu)Lu-PSMA) for prostate cancer (PCa). Lutetium-177’s β-particles induce DNA damage, leading to tumor cell death with minimal harm to healthy tissues, significantly improving progression-free survival and quality of life. However, further insights into local treatment effects are essential for optimization. To enhance understanding, we analyze the DNA damage response (DDR) and immune response induced by TRT in cell lines, ex vivo tumor slices, and xenografted mice using live-cell imaging, molecular biology, and histology. We have demonstrated diverse DNA damage patterns in tumor and normal tissue cells. Additionally, RNA expression analysis, drug screens, and CRISPR-Cas9 editing are helping elucidate the broader cellular responses. A collaborative project with Dr. Sophie Veldhuijzen van Zanten focuses on identifying TRT strategies for pediatric neuro-oncology. • Tumor radiobiology of NET TRT (Pleun Engbers, Giulia Tamborino, Tijmen de Wolf) • Tumor cell radiobiology of PCa TRT (Mariangela Sabatella, Txema Heredia Genestar) • Immune responses activated by TRT (Justine Perrin, Rob Verhagen) • Radiobiological assessment of blood of NET TRT patients (Nina Becx, Renata Brandt) • Novel TRT options for pediatric neuro oncology (Nina Overdevest, Txema Heredia Genestar) • Resistance mechanisms and heterogeneity in actinium-255 based TRT (Ana Gebara Garcia de Lima)

Accurate dosimetry remains a challenge in TRT. With Dr. Mark Konijnenberg, we are developing advanced dosimetric models to predict biological responses based on (micro)dosimetric quantities. These models are vital for understanding dose effects on diverse cellular targets and heterogeneous tumor regions, paving the way for their integration into TRT treatment planning systems. • Live cell imaging of DNA repair dynamics by TRT (Pleun Engbers, Tijmen de Wolf, Justine Perrin, Rob Verhagen) • Radiobiological comparison of lutetium-177 and actinium-225 for PCa TRT (Mariangela Sabatella) • Radiobiological comparison of lutetium-177 and terbium-161 (Mariangela Sabatella, Guillermo Garrido Hernandez) • Micro- and macrodosimetry of TRT (Giulia Tamborino) • Automated image analysis methods for TRT fluorescent microscopy (Tijmen de Wolf, Guillermo Garrido Hernandez) • Radiobiological comparison of holmium-166 and yttrium-90 for radioembolization (Justine Perrin) • Radiobiological comparison of non-targeted radionuclides in 2D systems (Duschka Kleijn)

Figure 1. Different types of radionuclides being used in our research and a schematic scheme on DNA damage induction.

Radiobiology and dosimetry of different radiation qualities Beyond lutetium-177, we study other radionuclides such as terbium-161 (β- and conversion electron emitter), actinium-225 (α-emitter), holmium-166, and yttrium-90 (both β-emitters). These radionuclides vary in decay type, half-life, and range, influencing their cellular effects. In collaboration with Dr. Erik de Blois, we compare their radiobiological effects through in vitro studies and in silico dose simulations to match radionuclides with clinical indications effectively. Figure 2. Analysis of the dose-effect relationship. Linking biological effects to absorbed dose calculated using dosimetric simulations.

106

scientific report 2025 | RADIOBIOLOGY OF RADIONUCLIDE THERAPY

Radiosensitization to improve radionuclide therapy outcome Our work highlights that TRT efficacy can be enhanced by combining it with radiosensitizing compounds. Specifically, DDR inhibitors impair DNA repair mechanisms, amplifying TRT-induced cell death. For instance, combining TRT with the PARP-1 inhibitor olaparib for NET tumors has led to clinical trials globally, including our phase I trial in collaboration with Dr. Hans Hofland. We also explore additional radiosensitizers, such as DNA-PKcs and HSP90 inhibitors. DNA-PKcs inhibition has shown a significant increase in therapeutic efficacy without detectable toxicity in preclinical models. We are now investigating the mechanisms behind HSP90’s radiosensitizing properties to refine these combinatory regimens further. • Preclinical radiosensitization to improve TRT outcome (Mariangela Sabatella, Renata Brandt, Txema Heredia Genestar) • Clinical phase 1 trial of NET TRT in combination with PARP inhibitors (Nina Becx, Renata Brandt, Txema Heredia Genestar)

Nonnekens, Julie ERC starting grant 2021: ‘RADIOBIO: Deciphering the radiobiology of targeted radionuclide therapy: from subcellular to intra-tumoural analyses’. 20222027 Veldhuijzen van Zanten, Sophie, and Julie Nonnekens Research grant Cure Starts Now Foundation: ‘Development and optimization of targeted radiopharmaceutical therapies for pediatric brain tumors; a world-first translational study’. 2023-2027 Perrin, Justine MSCA postdoctoral fellowship: ‘Impact of BRCA2 deficiency on the DNA damage response and immunogenicity of prostate cancer after radioligand therapy’. 2024-2026 Nijsen, Frank, Julie Nonnekens, Sandra Heskamp, Antonia Denkova, and consortium partners NWO Perspectief consortium: ‘Understanding the radiobiology of therapeutic medical radionuclides’. 2023-2028

Invited Lectures Julie Nonnekens. ‘Radiobiology in radionuclide therapy: key to understanding dose-effect relationships’. National Cancer Institute USA, online. Jan 2025.

Expectations & Directions Our team combines cutting-edge technological and radiobiological expertise to enable the clinical implementation of optimized therapeutic strategies. Our research aims to deepen understanding of the radiobiological effects of TRTs, a field with significant knowledge gaps. By uncovering the mechanisms underlying TRT-induced cellular effects, we aim to build a foundation for novel research opportunities, as current efforts only scratch the surface of TRT radiobiology. Future directions will expand on key areas, such as how the physical and biological properties of radiolabeled compounds influence radiation dose delivery and treatment outcomes. Additionally, we will explore the role of the tumor microenvironment and systemic responses in shaping TRT efficacy. These insights will not only improve our understanding of TRT but also facilitate the development of tailored therapeutic regimens, moving closer to personalized and curative cancer treatment.

Funding Kanaar, Ronald, Julie Nonnekens, Hans Hofland, Ferry Eskens, Wouter de Herder, Tessa Brabander, Astrid van der Veldt, Mark Konijnenberg, and Stijn Koolen Oncode clinical proof of concept study: ‘Improving Peptide Receptor Radionuclide Therapy with PARP inhibitors: the PRRT-PARPi study’. 2021-2026

Julie Nonnekens. ‘Enhancing lutetium-177 radionuclide therapy with DDR inhibitors’. American Association of Cancer Research annual congress, Chicago, USA. April 2025. Julie Nonnekens. ‘Introduction radiobiology and recommendations for reporting preclinical studies’. Netherlands Society for Clinical Radiochemistry meeting, Zevenaar, the Netherlands. June 2025. Julie Nonnekens. Expert during ‘challenge the expert’ session on clinical application of radiobiology for radionuclide therapy. European Association of Nuclear Medicine (EANM) annual congress, Hamburg, Germany. Oct 2025. Julie Nonnekens. ‘Choosing the right radiation modality: Radiobiological insights into alpha, beta & external beam radiation’. Targeted Radiopharmaceutical Summit, Amsterdam, the Netherlands. Nov 2025. Julie Nonnekens. Keynote lecture: ‘Basics in radiobiology & dosimetry, lessons from external beam and radionuclide therapy’. Netherlands Society of Nuclear Medicine, Utrecht, the Netherlands. Nov 2025. Julie Nonnekens. ‘Radiobiological comparison of different radiation qualities: Alpha vs beta vs external beam radiation’. ESTRO annual congress, Vienna, Austria. May 2025.

107

Additional Personnel Renata Brandt – Research technician Rob Verhagen – Research technician Doris Houtzager – intern, 2nd year MSc student Nanobiology, Erasmus University and TU Delft. Sept 2024 – June 2025. Daily supervisor Justine Perrin. Stefan van Alen – intern, 2nd year MSc student Nanobiology, Erasmus University and TU Delft. Sept 2024 – June 2025. Daily supervisor Pleun Engbers and Tijmen de Wolf. Ana Gebara Garcia de Lima – intern, 2nd year MSc student Molecular medicine, Erasmus University. Sept 2024 – June 2025. Daily supervisor Mariangela Sabatella. Mimi Welte – intern, 3rd year BSc student University college, Erasmus University. Nov 2024 – May 2025. Daily supervisor Nina Overdevest. Luiza Fernandes Moreira - intern, 3rd year BSc student University college, Erasmus University. June 2025 - June 2026. Daily supervisor Duschka Kleijn.

Post-docs

Giulia Tamborino, PhD Cellular Radiation exposure effects of molecular radioNuclide Therapies TRT offers a promising treatment for systemic cancers, but its effectiveness can be enhanced by tailoring treatment to individual patient needs. Dosimetry, or radiation dose calculations, plays a crucial role in optimizing radiation-based treatments by customizing regimens to reduce toxicity and improve tumor responses. However, TRT’s radiobiological mechanisms differ from traditional external beam radiotherapy (EBRT), particularly at low absorbed doses or low dose rates. Understanding these differences at the cellular and molecular levels is essential for advancing TRT.

To improve current dosimetry methods, we aim to develop accurate computational models that correlate microdosimetry with biological effects in both in vitro and in vivo settings. The creation of such computational frameworks and the investigation of the radiobiology of dose-response in TRT in cells and in small animals can ultimately lead to a better understanding of this treatment modality and to increase the predictive power of dosimetry.

108

scientific report 2025 | RADIOBIOLOGY OF RADIONUCLIDE THERAPY

Guillermo Garrido Hernandez, PhD Dosimetry and dose-effect modeling in targeted radionuclide therapy Targeted radionuclide therapy (TRT) targets tumor tissue selectively and damages the DNA of cancer cells, leading to their death. Existing models used to estimate the biological effects of TRT are adapted from photonbased external beam radiotherapy (EBRT) and do not consider key parameters that have been shown to have an effect on TRT radiobiological response. Radiobiological response to ionizing radiation in TRT is often interpreted using dose-effect relationships established for EBRT. The translation of such concepts to TRT is non-trivial due to not only dose-rate effects, but also to differences in radiation quality, subcellular

energy depositions, and the heterogeneity of radiation source distributions in TRT. This highlights the need for the development of explicit TRT-specific approaches to dosimetry, dose-effect modelling, and treatment optimization, safety and effectiveness. My research aims to advance TRT-radionuclide specific models to accurately estimate TRT radiation effects in vitro and in vivo. Ultimately, this work could potentially result in novel TRT strategies for cancer treatment.

Justine Perrin, PhD Impact of BRCA2 deficiency in TRT of prostate cancer

TRT has shown promising results in the treatment of metastasized prostate cancer. Prostate specific antigen (PSMA) overexpression in these cancers enables targeted therapy using a lutetium-177 radiolabeled peptide. The VISION phase III trial showed notable improvement in overall survival compared to standard care.

My research investigates the role of BRCA2 deficiency, a common feature in the most aggressive phenotype of prostate cancer. BRCA2 is critical for homologous recombination repair of DNA double-strand breaks, and its absence may serve as a biomarker for increased sensitivity to TRT.

However, not all patients respond to TRT, highlighting the need for prediction of good responders early to avoid unnecessary side effects.

My research focus on the impact of BRCA2 deficiency on TRT efficacy in prostate cancer cell lines, to identify if this deficiency makes cells more sensitive to TRT, and how it impacts the immunogenicity.

109

Mariangela Sabatella, PhD Radiobiology of alpha and beta-PSMA targeted radionuclide therapy TRT uses radionuclides (β- or α-emitters) linked to tumorspecific inhibitors. In metastasized prostate cancer (mPCa), prostate-specific membrane antigen (PSMA) is the main target. β-emitter-based PSMA-TRT has shown to improve overall survival in mPCa patients. However, not all patients respond to treatment, and some experience severe side effects, highlighting the need for improved therapies and novel PSMA inhibitors. Recently, α-emitters have gained attention due to their ability to cause more complex and irreparable DNA damage, potentially enhancing therapeutic efficacy. Despite these promising characteristics, limited

knowledge exists about the type of DNA damage induced by β- or α-emitters and the DNA damage response (DDR) mechanisms that tumor cells activate to survive the therapy. My research focuses on unraveling the molecular mechanisms underlying the effects of β- and α-emitters in PSMA-TRT and identifying DDR targets. This knowledge may lead to combination therapies with DDR inhibitors, improving outcomes for mPCa patients.

Txema Heredia Genestar, PhD Leveraging bioinformatics to understand the response to TRT

TRT is a promising treatment for patients with metastasized cancers. TRT comes in many forms, with differences in radioligand, isotope, and type-andenergy of the emitted radiation. The project aims to systematically characterize the effects of different radionuclides across different tissues and tumor types. Multiple Dutch institutes collaborate in the study of TRT’s effects in cell cultures, organoids, and the interactions with the immune system in in vivo studies.

respond to therapy short after treatment and how the different radiation types affect the cellular repair pathways involved in overcoming DNA damage. When observing cells under the microscope, we can only look at a limited number of proteins at once. However, bioinformatic tools allow us to explore the larger transcriptomic response, in hope to find targets that will improve therapy.

Our group focuses on the in vitro effects of TRT in cell culture as well as the analyses of cellular response using bioinformatic tools. We aim to explore how cells

110

scientific report 2025 | RADIOBIOLOGY OF RADIONUCLIDE THERAPY

PhD Students Ana Gebara Garcia de Lima, MSc

Duschka Kleijn, MSc Advisors Julie Nonnekens, Roland Kanaar & Justine Perrin

Advisors Julie Nonnekens, Mariangela Sabatella

Investigating incomplete responses to actinium-225-based TRT: mechanisms and cellular heterogeneity

In vitro 2D radiobiological effects of radionuclides with different physical properties

This project investigates the impact of target protein expression heterogeneity in hepatocellular carcinoma (HCC) cells on TRT efficacy. Radiobiological effects are measured upon treatment with TRT, and recurrent cells will be re-treated and re-characterized to better understand mechanisms for treatment resistance.

In this project, we aim to map the direct radiobiological response of various liver and prostate cancer cells to a set of radionuclides in vitro. We will specifically examine tumor cell survival, DNA damage repair capacity, and the activation of specific cellular pathways.

Pleun Engbers, MSc

Tijmen de Wolf, MSc

Advisors Julie Nonnekens & Mariangela Sabatella

Advisors Julie Nonnekens, Ihor Smal & Roland Kanaar

Image analysis for PRRT

Subcellular & intra-tumoural radiobiology of TRT

This project aims to understand the radiobiology of peptide receptor radionuclide therapy (PRRT). Live cell microscopy is used to capture the dynamics of the radionuclides and DNA damage over time. Image analysis is an essential step required for the quantification. We develop novel algorithms to improve and automate the analysis.

In this project, we aim to identify biological parameters that are important in the radiation effect of TRT. In addition, we perform live cell imaging in vitro and in vivo, to determine subcellular and intratumoral localiziation of 177Lu-DOTATATE. Finally, we aim to evaluate DNA damage induction and repair kinetics after TRT.

Nina Becx, MSc Advisors Julie Nonnekens, Hans Hofland & Roland Kanaar

Improving Peptide Receptor Radionuclide Therapy with PARP inhibitors Peptide receptor radionuclide therapy (PRRT) with the beta-emitting radiopharmaceutical 177LuDOTATATE is an effective and safe treatment option for patients with metastatic neuroendocrine tumors (NETs). Response rates are still limited, so there is a need for improvement. The PARP inhibitor olaparib has the potential to improve this treatment because of its radiosensitizing effects.

111

Simone Dalm, PhD associate professor 112

RADIOTRACER INTERACTIONS

Top Publications 2025

Simone Dalm has a BSc in Biomedical science and a MSc in Oncology, which she obtained at the VU University in Amsterdam. In 2012 she started her PhD at the department of Nuclear Medicine of the Erasmus MC and in 2017 she graduated cum laude. She then continued with a postdoc at the same department (by that time the department of Radiology and Nuclear Medicine had joined forces). During her postdoc she established her own research line and started her research group: The Radiotracer Interactions Group. In September 2019 she was promoted to assistant professor and in December 2025 she became associate professor.

Van der Heide CD, R McMorrow, F Gama-Franceschi, M Doukas, L Mezzanotte, S Dalm. Illuminating radionuclide therapy response: Monitoring cancer cel land CAF survival in a direct 2D co-culture model. Nucl Med Biol. 2025; 150-151:109571. Damiana TS, L vam den Brink, LW de Kreij-de Bruin, DC Stuurman, CM de Ridder, S Dalm. Chemotherapy alters radiosensitivity and GRPR expression of prostate and breast cancer cells. EJNMMI Res. 2025;15:98.

She received multiple personal and industry sponsored grants (e.g. KWF Young Investigator’s Grant and ZonMw Veni) and is involved in several national and international research projects. She also received multiple awards for her achievements such as the Editors’ choice award of the Society of Nuclear Medicine and Molecular Imaging, multiple Alavi Mandell awards and the Research Prize of the Erasmus University. Her scientific interests include molecular biology, targeted therapy, and nuclear imaging and therapy.

Morgat C, H Duan, S Dalm, E Hindié, T Günther, BJ Krause, V Kramer, F Cavelier, AW Stephens, S Moran, L Lamb, A Lagaru. A vision for Gastrin-Releasing Pepride Receptor Targeting for Imaging and Therapy: Perspective from Academia and Industry. J Nucl Med. 2025; 66:1160-1167.

s.dalm@erasmusmc.nl

Context

arget-mediated radionuclide imaging and treatment are anti-cancer interventions successfully applied in the clinic. For this, radiotracers are applied directed against molecules that are overexpressed on tumor cells. Depending on the radionuclide connected, the same tracer can be used for imaging (ß+ and γ-emitting radionuclides) and treatment (ß- and α-emitting radionuclides). Our studies focus on (but are not limited to) the application of radiotracers targeting the gastrin releasing peptide receptor (GRPR) (overexpressed on e.g. prostate- and breast cancer), the prostate specific membrane antigen (PSMA) (overexpressed on e.g. prostate cancer and the neovasculature of various solid tumors), the somatostatin receptor subtype 2 (SSTR2) (overexpressed on neuroendocrine- and breast cancer) and fibroblast activation protein (FAP) expressed on cancer associated fibroblasts (CAFs). The latter are present in the tumor stroma of ~90% of all solid cancers. Our objective is to develop and evaluate novel radiotracers and application strategies, and to optimize the use of radiotracers in order to achieve more cure, fewer side effects and a better quality of life for cancer patients. This includes studies to identify patient groups best suited for application of a specific radiotracer, the development and the application of novel strategies to improve tumor targeting and minimize off-target organ toxicity, and studies to better understand the mechanism of action of radiotracers. To achieve this, we unravel the interaction of radiotracers with the human body on a molecular, cellular, tissue, organ, organ-system and patient level.

113

Research Projects: Objectives & Achievements

properties of prostate cancer (Tyrillshall Damiana) – Discovering new targets for radiotheranostics of neuroendocrine prostate cancer (Lisa Bokhout) – Using functional MRI for early detection of response to PSMA-targeted radionuclide treatment (in collaboration with Dr. Warnert)

Personalized treatment For targeted radionuclide imaging and therapy to be successful high expression level of the target in the tumor is a prerequisite. Next to this, whether the target is expressed by the tumor cells themselves or by surrounding cells present in the tumor microenvironment, and how target expression is distributed (heterogeneous vs homogeneous), are also important aspects determining the success of these interventions. Furthermore, during cancer development and progression the expression of the target can vary, and additionally, different cancer subtypes can have different molecular characteristics, including different target expression levels. Additionally, it is most probable that novel developed targeted radionuclide treatments will be applied in late stage diseases. In this case patients have often been treated with other types of anti-cancer treatments such as hormone therapy, chemotherapy or a combination of the two. These treatments can influence target expression as well as radiosensitivity of cancer cells, the latter being another key component for successful radionuclide treatment. In our studies, we evaluate the expression level and distribution of novel targets in healthy tissues and during the development and progression of cancer. In addition, we also study how other anti-cancer drugs can affect target expression and radiosensitivity of cancer cells. Ultimately our goal is to identify specific patient groups, with respect to disease stage, disease subtype and treatment history, that can benefit from specific targeted radionuclide imaging and treatment strategies. See Figure 1.

Figure 1. Target expression and radiosensitivity can vary between different disease stages and as a consequence of prior treatment. Identification of patient groups best suited for specific target-mediated imaging and therapy strategies is essential for the success of these interventions.

Novel therapeutic strategies, combination treatment and the tumor stroma Although targeted radiotracers are successfully applied clinically, complete response in patients is rare. Moreover, since healthy organs can also express the respective target and because of the physiological clearance of the radiotracers, healthy organs are often also exposed to radiation which can cause side effects. The above indicates that there is a need for novel developments to increase the efficacy and safety of targeted radionuclide treatment. In line with this, our focus is to develop novel and improved therapeutic strategies for targeted radionuclide treatment of cancer. This includes the development of novel radiotracers (Figure 2), the development of methods to improve tumor-to-background ratio of radiotracer uptake, targeting the tumor stroma, combining radionuclide therapy with other anti-cancer therapies or with strategies to sensitize cancer cells to radionuclide therapy, and tandem radionuclide/radiotracer treatment strategies. Projects: – Development and evaluation of NTS1-targeting radiopharmaceuticals (Tyrillshall Damiana) – Targeting FAP-expressing CAFs in the tumor stroma for radionuclide theranostics (Circe van der Heide) – Targeting proteoglycans with radiolabeled cellpenetrating peptides: a novel approach for detection and treatment of cancer (Joana Campeiro) – Tandem GRPR- and PSMA-targeting radionuclide treatment for prostate cancer management (Lisa Bokhout) – Novel biomarkers for targeting neuroendocrine prostate cancer (Lisa Bokhout) – Novel developments for GRPR-targeted theranostic interventions (in collaboration with Dr. Seimbille and the department of Medical Oncology) – Targeted radionuclide theranostics for sarcomas (in collaboration with Dr. Veldhuijzen van Zanten and the department of Radiology & Nuclear Medicine of the LUMC) (Lisette de Kreij-de Bruin)

Projects: – The effect of treatment history on the success of GRPRdirected radionuclide therapy (Tyrillshall Damiana) – GRPR expression in relation to pathological and clinical

114

scientific report 2025 | RADIOTRACER INTERACTIONS

Figure 2. Targeting highly sulfated heparan sulphate proteoglycans (HSPGs) on triple negative breast cancer (TNBC). Compared to healthy tissues and other breast cancer subtypes, TNBCs have an increased expression of highly sulphated HSPGs on their cell surface. We aim to exploit this aberrant expression pattern to develop a targeted strategy for radionuclide imaging and treatment of TNBC. Crotamine, a cell penetrating peptide, has shown high affinity to highly sulfated HSPGs and therefore provides the ideal basis for exploring this strategy. ER=estrogen receptor, PR=progesterone receptor, HER2=human epidermal growth factor receptor 2.

Understanding the mechanism of action of radiotracers In another attempt to improve the efficacy and safety of targeted radionuclide therapy, we aim to gain more understanding of the mechanisms of action of radiotracers. The obtained knowledge is subsequently used to develop novel strategies to improve the therapeutic index of radionuclide therapy. Examples include comparing radiotracers with agonistic and antagonistic properties, studying the difference in binding of radiotracers to cancer cells vs binding to healthy cells with physiological expression of the target, developing more accurate models for studying radiotracers (e.g. 2D vs 3D models) (Figure 3), and studying the effect of radiotracer treatment on the immune system and cancer immunity. The ultimate goal of these studies is to use the obtained knowledge to develop strategies to improve targeted radionuclide imaging and treatment.

Figure 3. 2D vs 3D cell culture for studying radiotracers. Amongst other factors, the increased cell-to-cell interactions and the recapitulation of the tumor microenvironment in 3D models offers a more representative and accurate system for radiotracer evaluation.

Expectations & Directions Our ultimate goal is to develop and improve novel therapeutic strategies using targeted radiotracers in a personalized setting to achieve more cure, fewer side effects and a better quality of life for cancer patients (Figure 4). We aim to achieve this by introducing novel radiotracers and application strategies for cancer imaging and treatment into the clinic. This also includes combination with other anti-cancer treatments and combinations of radionuclides. In addition, we will keep performing studies to better understand the interaction between radiotracers and their target at different levels of organization in the human body. This will provide opportunities to develop and apply novel radiotracers and application strategies.

Projects: – Development of a clinically representative co-culture model for FAP-targeted radionuclide therapy (Circe van der Heide) – 2D vs 3D models for preclinical radiopharmaceutical evaluation (Ilva Klomp) – The role of the tumor microenvironment in targeted radionuclide therapy responses (Ilva Klomp) – 3D prostate cancer model development for evaluating radiotheranostics in heterogeneous target expressing models (Lisa Bokhout) – In depth radiobiology of PSMA-targeted radionuclide therapy in 2D and 3D models for developing advanced dosimetry models (Consortium project, PhD student to be hired) Figure 4. Overview of research goals.

115

Funding

PhD Students

Klomp, Ilva NWO ENW XS: ‘Identifying the tumor stroma as a key player in resistance to internal radiation treatment’. 2024-2025 Campeiro, Joana NWO ENW XS: ‘Tackling triplenegative breast cancer: Development of a novel nuclear medicinebased strategy to create a personalized medicine approach for a biomarker-negative cancer type’. 20242025 Dalm, Simone Daniel den Hoed Award: ‘Taking Prostate Cancer Theranostics To The Next Level: PSMA- and GRPRtargeted Tandem Radionuclide Therapy for More Cure and Less Side Effects’. 2022-2026 Dalm, Simone Commercial collaboration Ratio Therapeutics: ‘Long circulating FAP tracers’. 2022-2025

Circe van der Heide, MSc Advisors Simone Dalm & Frederik Verburg

FAP- targeted radionuclide therapy Fibroblast activation protein (FAP) is expressed by cancerassociated fibroblasts (CAFs) present in the stroma of 90% of solid tumors, making it a relevant biomarker for targeted interventions, with potential pan-cancer application. The aim of our project is to increase understanding of the interaction between FAP-targeted radiotracers, CAFs, and cancer cells, and to develop clinically relevant models to study and evaluate (novel) FAP-TRT strategies.

Additional Personnel Lisette W. de Kreij-de Bruin – MSc, Research Technician Debra Stuurman – Biotechnician Corrina de Ridder – Biotechnician Indy Teijl - Intern Julian Knol - Intern

Post-docs

Joana Campeiro, PhD Targeting proteoglycans with radiolabeled cellpenetrating peptides: a novel approach for detection and treatment of cancer

Triple-negative breast cancer (TNBC) is the most malignant breast cancer subtype with (extremely) poor prognosis, despite treatment options such as surgery and chemo/radiotherapy. Proteoglycans (PGs) have shown to play a role in cell proliferation and invasive growth in different tumor entities, and their dysregulated and overexpression is associated with poor prognosis in TNBC. Currently, few anti-cancer therapies targeting PGs have been developed using antibodies or inhibitors, but the clinical translation of these compounds remains unsuccessful. Alternatively, we propose to exploit PGs as a target to actively and specifically deliver imaging and/ or cytotoxic radionuclides to TNBC cells for targeted radionuclide theranostic (TRT) purposes. Cell-penetrating peptides (CPPs) have a high affinity for PGs expressed on the cancer cell membrane, making them ideal candidates for the proposed PG-targeting interventions. Our aim is not only to generate and evaluate the first class of radiolabeled CPPs for PG-targeted TRT, but also to use the developed molecules to get a better understanding of the role of PGs in TNBC. Moreover, PGs are overexpressed on multiple other cancer types as well, and thus the proposed strategy can be beneficial for a wide range of cancers.

116

scientific report 2025 | RADIOTRACER INTERACTIONS

Ilva Klomp, PhD The role of the tumor microenvironment in targeted radionuclide therapy responses

Targeted radionuclide therapy (TRT) is clinically applied for the treatment of metastasized prostate cancer and neuroendocrine tumors. TRT has proven its clinical value, however, response rates are variable and a significant number of patients do not respond to TRT despite qualifing for treatment based on target expression observed on nuclear scans. Thus, there is a clear and urgent need for strategies to improve treatment outcomes. Next to target expression, the tumor cells’ radiosensitivity is essential for cancers’ response to TRT. Several studies have demonstrated that specific components of the tumor microenvironment (TME) can induce epithelial-tomesenchymal transition (EMT) in cancer cells. EMT is associated with an aggressive cancer cell phenotype, and, very important for TRT, radioresistance towards other forms of radiation therapy. The aim of our project is to unravel the exact role of EMT in TRT responses, with the ultimate goal of improving TRT outcomes by interfering with EMTinduced radioresistance. In addition, to adequately study the role of the TME in TRT response, we aim to develop clinically relevant threedimensional models and accompanying standard operation procedures for accurate analysis of the interaction between TRT, the TME and TRT response.

Tyrillshall Damiana, MSc

Lisa Bokhout, MSc Advisors

Advisors Simone Dalm & Frederik Verburg

Simone Dalm, Frederik Verburg & John Martens

Improving radiocuclide treatment for PCa

The effect of treatment history on the succes of targeted radionuclide therapy

Prostate cancer (PCa) treatment with prostate specific membrane antigen (PSMA)-targeting radiotracers is effective, but comes with side effects. Other promising radiotracers, with a more favorable safety profile, for PCa targeted radionuclide therapy (TRT) are gastrin-releasing peptide receptor (GRPR)-targeting radiotracers. However the prevalence of GRPR is lower tham that of PSMA. We therefore aim to improve TRT for PCa by studying the efficacy and safety of tandem PSMA- and GRPR-TRT by performing preclinical in votro an in vivo studies in clinically relevant models.

Gastrin-releasing peptide receptor (GRPR)-targeted radionuclide treatment (TRT) has shown promising results in preclinical and initial clinical studies. However, little is known aboout the influence of treatment history on the succes of GRPR-TRT. We aim to unravel the effect of priorly received chemotherapy and hormone therapy on GRPR-TRT, with a main focus on the effect on cancer cells’ radiosensitivity and target expression.

117

Erik Verburg, MD, PhD full professor 118

TRANSLATIONAL NUCLEAR MEDICINE

2022-present: Director, Cyclotron Rotterdam B.V.

Top Publications 2025

2020-present: Professor of Translational Nuclear Medicine and consultant in Nuclear Medicine, Erasmus MC.

van de Sanden JT, A Dusch, KM Westhoff, MR Bernsen, A Taubert, C Hermosilla, FA Verburg. Radiolabeling of Angiostrongylus vasorum- and Crenosoma striatum larvae: a novel method using PET/CT to unveil larval migration in the gastropod intermediate host (Lissachatina fulica). Parasit Vectors 2025; 18 :466.

2016-2020: Professor of Experimental Nuclear Medicine and deputy head of the Department of Nuclear Medicine, Philipps-University Marburg and University Hospital GieBen-Marburg, Marburg, Germany.

Coerts HI, TM van Ginhoven, WE Visser, AD Windhorst, MR Vriens, MJ van Treijen, B de Keizer, FA Verburg. [18F] Tetrafluoroborate PET/CT for detecting differentiated thyroid cancer: phase-II study protocol. Future Oncol. 2025; 21: 3167-3174.

2010-2016: Consultant in Nuclear Medicine and Assistant Professor of Nuclear Medicine, RWTH Aachen University Hospital, Aachen, Germany. 2008: PhD, Netherlands.

Utrecht

University,

Utrecht,

the

2005-2010: Training in nuclear medicine, University Medical Center Utrecht, Utrecht, the Netherlands, and University Hospital Würzburg, Würzburg, Germany. 1997-2004: Medical studies at the Catholic University Leuven, Leuven, Belgium, and Utrecht University, Utrecht, the Netherlands. f.verburg@erasmusmc.nl

Context

he discipline of nuclear medicine has a rich tradition of successful translational research at the Erasmus MC. Notable examples to come out of this tradition are the commercial imaging product Octreoscan® and more recently the radionuclide therapeutic agent Lutathera®. The preclinical research group of our department is world-renowned.

In order to facilitate and optimize the integration of new preclinical, technical and multimodality imaging concepts into clinical research and the transfer of these concepts from research to clinical practice, the Department of Radiology and Nuclear Medicine has newly created the Chair of Translational Nuclear Medicine in 2020. The year 2025 was characterized by many activities aimed at further building up the research line with numerous PhDs starting within the line and associated lines from assistant- and associate professors. In February 2024 we welcomed the new digital GE Omni Leged 32 cm standard axial field of view PET/CT scanner, followed in december by the installation of the first installation of the GE Omni 128cm long axial field of view PET/CT scanner in the northern hemisphere. These new digital PET/CT cameras offer completely new options for e.g. completely dynamic total body PET/CT imaging or ultralow activity imaging. The optimal use of this novel technology is an ongoing focus of research in Translational Nuclear Medicine. In 2026, the upgrade of the 32 cm axial field of view scanner to a 64 cm axial field of view length is planned, further developing the elongated axial field of view PET/CT concept.

119

Research Projects: Objectives & Achievements

It is expected to further expand the field of translational nuclear medicine in the near future:

Projects: • FDG-PET/CT for the detection of gastric cancer metastases (Erik Vegt)

•

The Chair of Translational Nuclear Medicine has driven the process for replacement of existing nuclear medicine cameras. Installation of a revolutionary new SPECT/CT was completed in 2023.

•

The Chair of Translational Nuclear Medicine was named Director of the Cyclotron Rotterdam BV in 2022; clinical and scientific integration of the Cyclotron Rotterdam B.V. and associated facilities within the Department of Radiology and Nuclear Medicine is being effectuated in the 2023-2027 timeframe.

•

Radiolabeling of inhaled pulmonary medication delivery systems for PET/CT-based airway deposition assessment (Jochem Wolfert)

•

A novel terrestrial molluscan model for medical imaging (Joep van de Sanden)

•

[18F] Tetrafluoroborate PET/CT scan for effectively detecting thyroid cancer metastases (Hannelore Coerts)

•

the European Thera4Care consortium, for research into novel concepts for radionuclide therapy.

•

Long axial field of view PET/CT: evaluation and clinical value.

Expectations & Directions Objectives •

Establish internal and external cooperation between Erasmus MC Academic Centers of Excellence, the TU Delft and clinical and preclinical groups with the Department of Radiology and Nuclear Medicine to further developments in tracer-based medical imaging and radionuclide therapy. Furthermore, establish connections with other early adopters of 128cm and 64 cm axial field of view PET/CT scanners.

•

Introduce novel tracers into clinical research and clinical practice for radionuclide imaging and therapy.

•

Introduce novel radionuclides into clinical research and clinical practice.

•

Establishment of novel imaging techniques such as PET/MRI, very high-resolution dynamic SPECT/CT and dynamic total body PET/CT in clinical research and clinical practice.

•

Establish novel non-vertebrate animal models to further improve the process of translational tracer development.

Expectations and directions In 2026 much effort will be made to establish the use of LAFOV PET/CT in clinical practice and scientific work. Furthermore, in 2026-2027, the upgrade of the current conventional length PET/CT scanner to an elongated 64 cm axial field of view is expected, which will also be associated with much work in order to establish the additional clinical value of this extended field of view in clinical practice and science.

Invited lectures Erik Verburg. ‘Nucleaire Geneeskunde: van Therapie naar Diagnostiek naar Theranostiek’. Bataafsch Genootschap voor Proefondervindelijke Wijsbegeerte, Rotterdam, the Netherlands. Jan 2025. Erik Verburg. ‘Theranostics at EMC’. Meeting of the Serbian Society of Nuclear Medicine, Novi Sad, Servia. April 2025.

Additional Personnel Marijn Mostert – Postdoc LAFOV PET/CT

Achievements •

The Chair of Translational Nuclear Medicine has driven the process for replacement of existing nuclear medicine cameras. In 2025, the installation of two new PET/CT scanners, one of which was a second-inthe world new model 128 cm long-axial-field-of-view scanner, was completed.

120

scientific report 2025 | TRANSLATIONAL NUCLEAR MEDICINE

Post-docs

Erik Vegt, MD, PhD Nuclear medicine diagnostics and therapy for gastric cancer and liver cancer Liver cancer and gastric cancer are the second and thirdleading causes of cancer-related death worldwide. In our research into these diseases, we closely collaborate with UMC Utrecht and Leiden UMC. Transarterial radioembolization (RE) is an effective treatment for locally advanced hepatocellular carcinoma (HCC), but optimizing treatment effect whle minimizing side effects is challenging. In the iHepar study, we are going to investigate the safety and efficacy of dosimetry-based individualized treatment planning of Ho-166 RE, hoping to maximize treatment effect, while minimizing toxicity. In addition, we are working on a novel preclinical model of RE, in whichh we are working on a novel preclinical model of RE, in whichh we are going to study the immune effects of RE and possibilities for combining RE with immunotherapy. For gastic cancer, the accuracy of gastroscopy and CT for detecting metastases Is limited. Thus, some patients Incorrectly undergo curative Intent treatment, with risk of complication and mortality. In our multicenter study “evaluation of FDG-PET/CT and LAparoscopy In STagIng advanced gastriC cancer (PLASTIC)”, we evaluate the value of FDG-PET/CT and SL In patients with locally advanced gastric canccer. First results show that FDG-PET/CT detected distant metastases In 3% of patients, and SL detected peritoneal or locally nonresectable disease In 19% of patients. We concluded that FDG-PET/CT has limited added value, but SL has considerable Impact. Data about quality of life and cost-effectiveness are still being collected and analysed. In addition, a follow-up study Is being planned, Investigating the value of the new PET tracer FAPI.

Mark Konijnenberg, PhD Interaction of dosimetry and radiobiology in targeted radionuclide therapy TRT New TRT modelities are increasingly being introduced for various metastatized forms of cancer. Personalisation of TRT Is sparcely performed by using dosimetry based on patient-specific uptake and distribution kinetics. Traditional TRT with beta-emitting radionculides like 131I and 177Lu can benefit from personalisation by maximising the absorbed dose to tumours while keeping the absorbed dose to organs at risk within acceptable dose constraints. These constraints are, however, not well known for TRT and were transposed from external beam radiotherapy experiencee. Radiobiology needs to be studied specifically for TRT, as radiation exposure in TRT is very different. DNA damage repair is taking place while the absorbed dose is still cumulating and dosimetry should take this specific radiobiology Into account when setting up calculational models. Specificly dosimetry for alpha-particel emitter 225Ac-PSMA has been developped to report its absorbed doses In the phase I clinical trials and preclinical experiments yield knowledge on the relative biological effectiveness of the alpha-emitter 225Ac in comparison to 177Lu and external beam radiotherapy. All projects are aimed at gathering data on dosimetry and TRT radiation effects in order to build treatment planning tools for TRT.

121

Ties Mulders, MD, PhD Project Translational Nuclear Medicine During the past year, my research activities focused on the optimization and clinical application of PET/CT imaging within our department. A major area of investigation involved the implementation and optimization of a newly installed PET/CT system, with particular emphasis on balancing radiation dose and acquisition time while maintaining diagnostic image quality. These projects aimed to improve patient safety, workflow efficiency, and quantitative imaging performance. In addition, I contributed to studies evaluating the added clinical value of PET/CT in patients with thymic epithelial tumors (thymomas), assessing its potential role in staging, treatment planning, and disease characterization. Furthermore, we initiated research exploring the integration of PET/CT into lung cancer screening pathways, focusing on its potential to enhance risk stratification and improve the diagnostic work-up of screen-detected pulmonary lesions. Collectively, these projects support the advancement of evidence-based molecular imaging and the optimization of patient care.

122

scientific report 2025 | INTERVENTIONAL RADIOLOGY RESEARCH

PhD Students Adnane Zerguit, MSc

Hannelore Coerts, BSc

Advisors Erik Verburg & Mark Konijnenberg

Advisors Erik Verburg, Tessa van Ginhoven, Bart de Keizer & Menno Vriens

Prediction of 225Ac dose–response based on 177Lu pharmacokinetics

[18F]Tetrafluoroborate PET/CT for Detection of Thyroid cancer

The project investigates the predictive value of 177Lu pharmacokinetics for 225Ac behavior in targeted alpha therapy. It combines quantitative imaging, preclinical dosimetry, and radiobiological analysis to study radioligand distribution, the impact of recoiling daughters, and differences in relative biological effectiveness between 225Ac- and 177Lu-based radiopharmaceuticals.

[18F]Tetrafluoroborate PET/CT has the potential to diagnose lymph node metastases and RAI-refractory differentiated thyroid cancer. Our aim is to introduce the [18F]Tetrafluoroborate PET/CT scan in the Netherlands, and to prospectively investigate whether this technque can detect thyroid cancer effectively.

Jochem Wolfert, PharmD Advisors Harry Hendrikse, Erik Verburg, Liesbeth Ruijgrok, Ties Mulders & Erik de Blois

Hidde IJtsma, BSc Advisors Erik Verburg, Tessa van Ginhoven, Menno Vriens & Bart de Keizer

Radiolabeling of orally inhaled drug products to assess lung deposition

A novel PET tracer for the detection of Differentiated Thyroid Carcinoma

We are aiming to develop novel radiolabeling methods to assess lung deposition of orally inhaled drug products using PET-CT. These imaging methods can then be employed to improve knowledge of treatments used in pulmonary disease.

18F-Tetrafluoroborate is a promising PET tracer targeting the sodium-iodide symporter expressed by thyroid (cancer) cells. This project aims to determine the optimal acquisition time point, evaluate whether metastatic disease to the lymph nodes can be detected and assess the correlation between tracer uptake in recurrent disease and 131I SPECT/CT findings.

Stanislav Ivanov, MSc Advisors Erik Verburg & Mark Konijnenberg

Joep van de Sanden, MSc Advisors Erik Verburg & Monique Bernsen

Radiolabeling of orally inhaled drug products to assess lung deposition We are aiming to develop novel radiolabeling methods to assess lung deposition of orally inhaled drug products using PET-CT. These imaging methods can then be employed to improve knowledge of treatments used in pulmonary disease.

Developing a new terrestrial molluscan model for medical imaging Setting up a facility for Achatina Fullica and Limax Maximus culturing. Determinating the function of Thyroid hormones in these 2 species. Assembling a whole genome sequence for both species. Using the medical imaging facilities to validate the use of snails and snugs as novel animal models in medical imaging.

123

Tessa Brabander, MD, PhD assistant professor

Astrid van der Veldt, MD, PhD associate professor 124

CLINICAL NUCLEAR MEDICINE: IMAGING AND THERAPY IN ONCOLOGY JOINT APPOINTMENT IN MEDICAL ONCOLOGY

Top Publications 2025

Astrid van der Veldt is a medical oncologist at Erasmus MC. After obtaining her cum laude medical degree, she completed two PhDs in medical oncology and nuclear medicine, resulting in a double dissertation with cum laude degree in 2012 at the VU University Medical Center in Amsterdam. In 2016, she completed her training in internal medicine and medical oncology at the VU University Medical Center and The Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital, respectively. Since 2017 she works at the Departments of Medical Oncology and Radiology & Nuclear Medicine at Erasmus MC. In clinical oncology, her main research interests include immunotherapy and the treatment of renal cell cancer and melanoma. In addition, she supervises research collaborations between medical oncology and imaging, including radiology and nuclear medicine.

Derks SH, LS Ho, SR Koene, MP Starmans, E Oomen-de Hoop, A Joosse, MJ de Jonge, KA Naipal, JL Jongen, MJ van den Bent, M Smits, AA van der Veldt. Size matters: Early progression of melanoma brain metastases after treatment with immune checkpoint inhibitors. Neurooncol Adv. 2025; 7:vdaf026. Ling SW, Q de Lussanet de la Sablonière, MAnanta, E de Blois, SLW Koolen, RC Drexhage, J Hofland, DGJ Robbrecht, AAM van der Veldt, FA Verburg, T Brabander. First real-world clinical experience with [177Lu] Lu-PSMA-I&T in patients with metastatic castration-resistant prostate cancer beyond VISION and TheraP criteria. Eur J Nucl Med Mol Imaging; 2025; 52:2034-2040. Cox CP, T Brabander, FA Verburg, M Segbers. Impact of patient size on image quality of OSEM3D and BSREM reconstructions in [68Ga]Ga-DOTA-TATE PET/MR. Cox CPW, M.EJNMMI Phys; 2025; 2:60.

a.vanderveldt@erasmusmc.nl

Tessa Brabander is a nuclear medicine physician and radiologist at the Erasmus MC. In 2011, she graduated in medicine at the Erasmus University. After graduation, she completed training in nuclear medicine in 2016 and started working as a staff member at the Department of Radiology & Nuclear Medicine. During her residency, she started research at the Department of Nuclear Medicine under supervision of prof dr DJ Kwekkeboom. In October 2017, she received her PhD. Currently, she works as a nuclear medicine physician and radiologist at the department. Her main research interests are imaging, therapy of neuroendocrine tumors, and other radionuclide therapy with alpha emitting radionuclides. t.brabander@erasmusmc.nl

Context

uclear medicine is rapidly changing as a result of the ongoing development of new radiopharmaceuticals. A variety of radiopharmaceuticals is currently available for imaging and their number is still increasing. In the last decade, many new tracers were developed for both diagnostics and radionuclide therapy. In clinical practice, nuclear medicine plays a pivotal role in the diagnosis of many diseases, within for example cardiology, neurology, and medical oncology. The concept of “theranostics” refers to compounds that can be used for both imaging and therapy. To this end, different radionuclides are labeled to the same compound. After injection, compounds labeled with a short-lived radionuclide (e.g. 68Gallium) provide the biodistribution of the compound in the body, which enables the prediction of clinical response to the same compound when labeled with a therapeutic radionuclide such as 177Lutetium (177Lu). Peptide receptor radionuclide therapy (PRRT) with [177Lu-DOTA,Tyr3]octreotate (177Lu-DOTATATE) is an example of a very successful theranostic, as this treatment improves progression-free survival in patients with advanced progressive, somatostatin receptor positive neuroendocrine tumors (NETs).

125

Research Projects: Objectives & Achievements

improve the number of somatostatin receptors in NETs. For therapy of prostate cancer, the peptide of first choice is PSMA. In addition, Erasmus MC is currently participating in a phase I clinical trial with 177Lu-NeoBomb, targeting GRPR, which is expressed in several cancers.

Projects: – The influence of new developments on the quality of PET imaging (Tiny Cox) – Peptide receptor radionuclide therapy using novel radionuclides for metastatic neuroendocrine tumor patients (Eline Zoetelief) – A tailored survivorship care plan for metastatic melanoma survivors (Imren Özdamar) – 225Ac-PSMA I&T for treatment of metastatic castration-resistant prostate cancer patients (Sui Ling) – Novel MRI techniques for the detection of brain metastases (Sophie Derks) – Gaining insight into the immune response to radium-223 therapy. (Swaaj Ling)

Diagnostics 18 F-FDG is the most well known PET tracer for clinical diagnostics. This tracer is frequently applied for staging in oncology, but it is also evaluated for response evaluation during therapy, which is illustrated by the IMPACT-study in colorectal cancer. As prostate cancer is not 18F-FDG avid, other tracers, like 68Ga-(prostate specific membrane antigen)PSMA, have been introduced for clinical evaluation of prostate cancer. As 68Ga-PMSA PET provides high image quality, it is increasingly applied for staging of prostate cancer. In 2021, 68Ga-PMSA PET was evaluated to guide bone biopsies in patients with metastatic prostate cancer. In addition, 68Ga-PMSA PET will also be investigated for response evaluation in patients with metastatic prostate cancer. In the IMPACT consortium (IMaging PAtients for Cancer drug selecTion), new tracers are evaluated in collaboration with VUmc, Radboudumc, and UMCG. In the IMPACT trial, the impact of PET using 18F-fluoroestradiol (18F-FES) and 89Zrtrastuzumab, is investigated for treatment planning of patients with metastatic breast cancer. In addition, PET using 89Zrgirentuximab, a monoclonal antibody targeting the glycoprotein CAIX, is evaluated during watchful waiting in patients with metastatic renal cell cancer.

Radionuclide therapy For decades, the Department of Nuclear Medicine has been focused on radionuclide therapy. Together with the NETTER-1 trial, the results of our large phase 2 trial has resulted in FDA and EMA approval of 177Lu-DOTATATE for gastroenteropancreactic neuroendocrine tumors (GEPNETs). Although 177Lu-DOTATATE has proven efficacy inpatients with NET, there is a need to further improve tumor response to PRRT. Therefore, we will investigate the combination of 177Lu-DOTATATE with different systemic therapies, e.g. chemotherapy or PARP-inhibitors (see chapter Julie Nonnekens). This may improve tumour response rates and hopefully progression free survival. In addition, epigenetic drugs are under investigation to

The department has invested in a new laboratory which is completely focused on the labeling of alpha emitting radionuclides. This laboratory has received a full GMP licence for the production of radiopharmaceuticals. These features are unique in Europe and will give us the opportunity to perform studies with alpha emitting radionuclides in patients. The first study that has started is the 225AcPSMA in a phase-1 clinical trial in patients with metastatic prostate cancer (see project Sui Wai Ling), funded by KWF/ Maarten van der Weijden foundation. Also, in 2021 a KWF Young Investigator grant was awarded to start with alpha emitting radionuclides in neuroendocrine tumor patients (225Ac-DOTATATE). Besides alpha emitters, the department also focuses on other new radionuclides for therapy. The wordwide shortages of Lutetetium-177 makes it necessary to search for new therapeutic options. One of these options is the use of Terbium-161, which has similar characteristics as Lutetium-177. The use of 161Tb-dotatate in patients will be made possible with the Erasmus MC fellowship.

Expectations & Directions To apply imaging from bench to bedside, collaborations with the preclinal imaging group will be further intensified. For diagnostics, future directions include the development and clinical implementation of new tracers. Using PET-MRI research projects between nuclear medicine and radiology (e.g. radiomics) were performed for e.g. prostate cancer (see project Sui Wai Ling) and melanoma brain metastases (see projects Sophie Derks and Imren Özdamar). The integrated PET-MRI combines a 3.0T MRI with the newest PET technology. This state-of-the-art scanner with simultaneous acquisition of PET and MRI creates many opportunities for imaging. The combination of these two techniques enables visualization of cellular changes by PET and localization by MRI. The PET-MRI can be used as a one-stop scan for staging of different tumor types. In the upcoming years, we will focus on new indications for PETMRI and its additional value compared to PET-CT or MRI alone.

Funding de Blois, Erik, Tessa Brabander Health Holland/TKI: ‘Phase 1 dose escalation study to evaluate tolerability and safety of Tb-PSMA in patients with metastatic, castration resistant prostate cancer’. 2024-2027 Brabander, Tessa, and Astrid van der Veldt KWF grant: ‘Phase I dose escalation study to evaluate tolerability and safety of 225Ac-PSMA in patients with metastatic prostate cancer’. 2019-2027

126

scientific report 2025 | CLINICAL NUCLEAR MEDICINE: IMAGING AND THERAPY IN ONCOLOGY

Van der Veldt, Astrid EMC fellowship: ‘Reducing toxicity and improving outcomes in immunotherapy treated melanoma patients’. 2019-2026

Tessa Brabander. ‘Diagnostics of neuroendocrine tumors’. GIO symposium YSL/SFG, Rotterdam, the Netherlands. Nov 2025.

Van der Veldt, Astrid DDH Award: ‘Early detecting and understanding treatment failure in melanoma brain metastases’. 2019-2027

Tessa Brabander. ‘Virtual NET tumor board Precision therapies and emerging strategies in NEN management’. NANETS-ENETS webinar, online. Dec 2025.

Van der Veldt, Astrid KWF Young Investigator Grant Bas Mulder Award: ‘Safe Stop-QoL: impact of early discontinuation of PD-1 blockade on quality of life (QoL) of patients with advanced melanoma’. 2019-2028

Astrid van der Veldt. ‘Update urological cancers’. PostASCO NVMO, Driebergen, the Netherlands. June 2025.

Van der Veldt, Astrid Trustfonds Erasmus: ‘Genomic landscape and actionable targets as identified by whole genome sequencing in metastases from patients with renal cell carcinoma’. 2021-2028 Brabander, Tessa KWF Young Investigator Grant: ‘Salvage therapy with 225Ac-DOTATATE for patients with metastatic neuroendocrine tumors’. 2022-2026 Brabander, Tessa Erasmus MC fellowship: ‘161Tbdotatate for neuroendocrine tumors’. 2023-2027 Van der Veldt, Astrid Transformation deal NFU: ‘Safe Stop IPI-NIVO trial: Early discontinuation of nivolumab upon achieving a (confirmed) complete or partial response in patients with irresectable stage III or metastatic melanoma treated with first-line ipilimumab-nivolumab’. 2022-2031 Van der Veldt, Astrid KWF grant: ‘In-depth study to provide safe long-term survivorship care to survivors of metastatic melanoma (SURVIVOR)’. 2023-2029 Van der Veldt, Astrid ZonMw Vidi grant: ‘Unravelling the tumour escape in melanoma survivors after stopping immunotherapy’. 2024-2029 Van der Veldt, Astrid (co-applicant) Research Synergy Grant: ‘Enhanced treatment and sustainable care: 3DPrinting of BRAF/MEK inhibitors’. 2024-2028

Astrid van der Veldt. ‘How can the immune system help to control cancer?’. European Conference on Interventional Oncology, Rotterdam, the Netherlands. April 2025.

Highlights Robert Stassen defended his thesis entitled ‘Overcoming boundaries in the prognostication of Melanoma: The role of the sentinel lymph node biopsy in an era of effective systemic therapies’

Additional Personnel Karlijn de Joode - Post-doc, Department of Medical Oncology Brigit van Dijk - PhD student, Department of Medical Oncology Joséphine Janssen - PhD student, Departments of Medical Oncology and Surgical Oncology Robert Stassen - PhD student, Department of Surgical Oncology Julia van den Broek - PhD student Department of Surgical Oncology

Invited Lectures Tessa Brabander. ‘Future perspectives of radionuclide therapy’. ECR annual meeting, Vienna, Austia. Feb 2025. Tessa Brabander. ‘Caution to mitigate toxicity in the optimization of RLT in real world patients’. ESMO sarcoma and rare cancers. Lugano, Switzerland. March 2025. Tessa Brabander. ‘Theranostics’. ECIO annual meeting. Rotterdam, the Netherlands. April 2025. Tessa Brabander. ‘SSTR imaging in GEPNET’. NVNG spring meeting, Bunnik, the Netherlands. June 2025.

127

PhD Students Sophie Derks, MD

Tiny Cox, BSc

Advisors Astrid van der Veldt, Martin van den Bent & Marion Smits

Advisors Erik Verburg, Tessa Brabander, Mark Konijnenberg & Marcel Segbers

PhD obtained: 24-06-2025 Preparing for PET PET image quality is of key improtance for optimal lesion detectability and interpretation of scans. Expansion of knowledge is requested about influence on image quality of new developments such as: new tracers, adapted patient preperation, digital PET/MRI and reconstruction algorithms to optimize PET image quality with adapted protocols and dosage regimen.

Brain metastases in real-world practice Patients with brain metastases (BMs) often have a poor survival. However, novel systemic therapies (i.e. targeted therapies and immune checkpoint inhibitors) have improved the overall survival for patients with cancer. We will develop novel magnetic resonance imaging (MRI) techniques for the detection of BMs. Since patients with BMs are often excluded from benchmark trials, we investigate real-world cohorts of patients with BMs of melanoma, renal cell cancer and non-small cell lung cancer. This data can provide helpful new insights about the effect of daily clinical practice decisions for clinicians at the Erasmus MC.

Imren Özdamar, MD, MSc Advisors Astrid van der Veldt, Henk Verheul & Marjolein Luchtenberg

Sui Ling, MD, MSc

Long-term survivorship care for survivors of

Advisors

metastatic melanoma

Tessa Brabander, Astrid van der Veldt & Erik Verburg

The introduction of immune checkpoint inhibitors has significantly improved the survival of patients with metastatic melanoma. For a tailored survivorship care plan, we will investigate the quality of life of these melanoma survivors. To develop the optimal follow-up schedule for radiological imaging, we will analyze residual lesions on radiological imaging over time.

225

Ac-PSMA I&T in patients with metastatic

castration-resistant prostate cancer Preclinical and preliminary clinical studies have shown that actinium-225 prostate specific membrane antigen (225AcPSMA) appears to be a promising radiopharmaceutical for therapy of metastatic castration-resistant prostate cancer (mCRPC). Our phase 1 dose-escalation study has already started and is currently still ongoing.

Micha van der Lee, MD Advisors

Eline Zoetelief, MSc

Astrid van der Veldt, Karlijn de Joode & Reno Debets

Advisors Tessa Brabander, Hans Hofland & Erik Verburg

Individualized treatment of metastatatic melanoma Immune checkpoint inhibitors (ICIs) have converted metastatic melanoma from a historically incurable into a potentially curable disease. However, a substantial number of patients experience an unexpected tumour escape after treatment with ICIs. We aim to early identify patients with metastatic melanoma who have progressive disease after immunotherapy, and to develop new treatment strategies for these patients. Therefore, we will characterize the macro- and microenvironment of patients with an unexpected tumour escape after discontinuation of immunotherapy.

Novel radionuclides for peptide receptor radionuclide therapy Improve and secure peptide receptor radionuclide therapy (PRRT) for patients with metastatic tumors (NETs) with new radiopharmaceuticals 225Actinium(alfa-emitting) or 161Tb- ( beta-emmiting ) DOTATyr3, octreotate (225Ac-/161Tb-DOTATATE). For both radiopharmaceuticals the tolerability and saftey are clinically evaluated in metastatic NET patients.

128

scientific report 2025 | CLINICAL NUCLEAR MEDICINE: IMAGING AND THERAPY IN ONCOLOGY

129

Yann Seimbille, PhD associate professor 130

RADIOPHARMACEUTICAL CHEMISTRY

Top Publications 2025

Yann Seimbille obtained his PhD in Radiopharmaceutical Sciences from the University of Sherbrooke under the supervision of Professors Johan van Lier and François Bénard. Following a postdoctoral fellowship at the University of California Los Angeles (UCLA) in the laboratories of Professors Daniel Silverman and Johannes Czernin, he assumed the role of Assistant Professor in the Department of Molecular & Medical Pharmacology. Subsequently, he joined the division of Nuclear Medicine & MolecularImaging at the University of Geneva and spent two years at Canada’s particle accelerator centre (TRIUMF) in Vancouver before joining Erasmus MC in July 2017. His research group is dedicated to advancing theranostics and multimodality imaging probes. His translational research program relies on cutting-edge (radio) chemistry to promote the application of functional imaging and targeted (radio)therapy in biomedical sciences.

Ma H, L de Muynck, R Houvast, S Beekman, A Piet, T March, L Hawinkels, S Mieog, A Vahrmeijer, Y Seimbille. Revolutionizing solid tumor surgery with fibroblast activation protein (FAP)-targeted imaging probes for a fluorescence-guided surgery of pancreatic cancer. Bioconjugate Chemistry 2025; 36:2145-2157. Hooijman E, J de Jong, C Ntihabose, F Bruchertseifer, A Morgenstern, Y Seimbille, T Brabander, S Koolen, E de Blois. Ac-225 radiochemistry through the lens of [225Ac]Ac-DOTA-TATE. EJNMMI Radiopharmacy and Chemistry 2025; 10, 9. Ntihabose C, M Handula, A Piet, S Beekman, L van Dalen, N Eskandari, A Aydogan, D Stuurman, C de Ridder, M Konijnenberg, Y Seimbille, E de Blois. Potentials and practical challenges of terbium-161 labeled radiopharmaceuticals. EJNMMI Radiopharmacy and Chemistry 2025; 10, 64.

y.seimbille@erasmusmc.nl

Context

he RadioPharmaceutical Chemistry group is actively engaged in the advancement of radiopharmaceuticals tailored for the diagnosis and treatment of cancer. Diagnostic radiopharmaceuticals are commonly administered to patients, and their distribution in the body is tracked using specialized imaging techniques, such as positron emission tomography (PET) or single-photon emission computed tomography (SPECT). It enables the non-invasive visualization and identification of pathophysiological processes. Therapeutic radiopharmaceuticals are designed to deliver targeted radiation dose to specific tissues or cells within the body, while minimizing damage to healthy organs.

Our research program is centered on theranostics and multimodality imaging probes, with a key focus on the clinical translation of these compounds. Design of these radiopharmaceuticals involves considerations of the chemical and biological properties of the compounds (e.g., binding affinity, specificity, stability, pharmacokinetics), as well as the characteristics of the radionuclide used (e.g., half-life, decay properties), to warrant effectiveness and safety. The following provides a concise overview of our ongoing research endeavors.

131

Research Projects: Objectives & Achievements Theranostics (Dx + Tx) Peptides and small molecules are highly attractive carriers for radiopharmaceuticals, owing to their rapid pharmacokinetics, high affinity, and effective tissue penetration. Recently, we developed innovative compounds that target specific biomarkers, including somatostatin receptor subtype 2 (SSTR2), gastrinreleasing peptide receptor (GRPR), prostate-specific membrane antigen (PSMA), fibroblast activation protein (FAP), and chemokine receptor type 4 (CXCR4) (Figure 1).

radionuclides (212Pb, 225Ac). Our recent work involved the labeling of somatostatin receptor subtype 2 agonists and antagonists with lead-212 and actinium-225, respectively. Comprehensive biological studies were conducted to assess the impact of chemical modifications and metal complexation on the binding properties of the peptides to SSTR2. Subsequently, we conducted evaluations on stability, biodistribution, pharmacokinetics, and dosimetry in tumor-bearing mice. While [225Ac]Ac-DOTA-JR11 showed a higher absorbed dose in the kidneys compared to [177Lu] Lu-DOTA-JR11, which may limit further clinical studies with this Ac-225 labeled peptide, our ligand labeled with 212, eSOMA-01, showed great potential for targeted alpha therapy of NETs. In fact, [212Pb]Pb-eSOMA-01 displayed a higher absorbed dose to the tumor, as well as a reduction in absorbed dose to the kidneys compared to the clinical reference DOTAM-TATE. Projects: – TAT of NET tumors (Amber Piet, Dylan Chapeau) – TAT of solid tumors (Lucas de Kruijff, Amber Piet) – Clinical implementation of Ac-225 labelled Somatostatin and PSMA radiopharmaceuticals (Eline Hooijman)

Theranostic Small-Molecule Drug Conjugates (T-SMDCs)

Figure 1. SPECT/CT images of T-47D tumor bearing mice after i.v. administration of: A) [111In]In-RM2; and B) [111In]In-hmdPP-02.

The amalgamation of these compounds into homo- or hetero-multimeric structures holds particular promise for enhancing tumor uptake and retention. Additionally, this approach addresses the intricate and diverse nature of primary tumors and metastases. Our ongoing preclinical assessments aim to pinpoint lead candidates for translation into clinical applications. Projects: – Development of theranostics for breast cancer (Priciana Paraïso) – FAP-targeted theranostic agents (Hanyue Ma, Le Li, Amber Piet) – Imaging and treatment of NET tumors with novel theranostics (Dylan Chapeau, Dylan Blijleven, Lucie Arnaud, Amber Piet)

Targeted Alpha Therapy (TAT) Recent clinical investigations have revealed a notable enhancement in the therapeutic efficacy of peptide receptor radionuclide therapy (PRRT) by substituting the conventional β-emitters (90Y, 177Lu) with alpha-emitting

The linkage of a cytotoxic drug to a targeting vector, such as an antibody, peptide, or small molecule, enhances the effectiveness of chemotherapy while minimizing toxicity to normal organs. Recently, we have developed compounds that target PSMA and FAP, coupled with mertansine (DM1). DM1 is commonly used in antibody-drug conjugates (ADCs), such as the approved T-DM1 (trastuzumab-emtansine). In contrast to ADCs, our theranostic small-molecule drug conjugates (T-SMDCs) also incorporate a chelator, enabling labeling with a gamma emitting radionuclide to produce a companion diagnostic or a therapeutic radionuclide for combined targeted chemotherapy and radionuclide therapy. Our studies indicated that our SSTR2-targeted T-SMDC, eSOMA-DM1 (Figure 2), showed promising characteristics in vitro, and tumor uptake in H69-xenografts higher at alltime points compared to [111In]In–DOTA-TATE. However, prolonged blood circulation led to increased accumulation of [111In]In–eSOMA-DM1 in highly vascularized tissues, such as the lungs, skin, and heart. Therefore, our results suggest the need for adjustments to optimize the distribution of eSOMA-DM1 prior performing therapeutic efficacy studies. Project: – Development of SMDCs and T-SMDCs targeting the tumor microenvironment (Hanyue Ma, Le Li, Amber Piet, Gijs Louwerens)

132

scientific report 2025 | RADIOPHARMACEUTICAL CHEMISTRY

Figure 2. Chemical structure of eSOMA-DM1. The SSTR2-targeting octreotate peptide (green) is conjugated to the cytotoxic drug DM1 (red) via a chelate-bridged linkage (blue). The drug is released by cleavage of the linker (black) by glutathione.

Fluorescence-Guided Surgery (FGS)

Clinical Radiochemistry

Recently, there has been an emergence of intraoperative imaging tools designed for real-time visualization of tumors. One such approach, termed fluorescenceguided surgery (FGS), has exhibited promising outcomes in aiding surgeons in accurately delineating tumor margins. This capability is particularly crucial for ensuring thorough tumor resection, reducing the risk of disease recurrence, and minimizing damage to healthy tissues. Despite these advancements, the availability of fluorescent imaging probes with optimal properties remains a challenge.

To address this gap, we synthesized fluorescent molecules targeting biomarkers (e.g., SSTR2, GRPR, FAP) prevalent in cancer. Notably, eFAP-24 demonstrated encouraging results in preclinical pancreatic tumor models (Figure 3). Project: – FAP-targeted imaging probes for precision fluorescence-guided surgery (Hanyue Ma, Amber Piet)

Radionuclide therapy is gaining considerable attention, and traditionally, the theranostic paradigm is using a molecule (e.g., DOTA-TATE) labeled with a gamma-emitting radionuclide (e.g., 68Ga) for imaging, and subsequently with a therapeutic radionuclide (e.g., 177Lu) for PRRT. However, the availability of the main therapeutic radionuclide, 177 Lu, is limited. This necessitates exploring alternatives, but the chemical and physical properties of these new radionuclides differ. It is, therefore, crucial to assess how labeling with these new radionuclides (e.g., 161Tb) impacts radiolabeling efficiency, stability, and the biodistribution of the radiotracer. Projects: – Radiolabeling of orally inhaled drug products to assess lung deposition (Jochem Wofert) – Tb-161 labelled Somatostatin and PSMA radiopharmaceuticals (Carolline Ntihabose) – The ‘Plug-and-Play Radionuclide Generator’ (Erik de Blois) – Ho-166 radiopharmaceuticals (Carolline Ntihabose)

Expectations & Directions

Figure 3. NIRF-color merge and NIRF images of tumor bearing mice at 8 h post-injection of eFAP-24. Images were captured using the clinical Artemis NIRF imager. White arrows indicate tumor localization.

The increasing prominence of radiopharmaceuticals in personalized medicine is anticipated to lead to a continued surge in their clinical utilization for both diagnosis and therapy. There is a growing demand for novel radiopharmaceuticals in research, driven by the need to deepen our understanding of human disease biology, improve diagnostic and therapeutic capabilities, and facilitate drug development. This heightened interest in radiopharmaceuticals motivates our group to actively pursue the development of new theranostics and imaging probes, with the potential to significantly impact the management of cancer patients.

133

Funding Seimbille, Yann, Carolien van Deurzen, and Agnes Jager Erasmus MC grants: ‘Theranostics hitting the Achilles’ heels of breast cancer: pointing the arrows at HER2 and GRPR’. 2021-2025 Ma, Hanyue Erasmus MC Research Innovation Grant: ‘A universal approach to image and treat cancer with novel promising fibroblast activation protein inhibitors’. 20232024 De Blois, Erik TKI Health Holland grant: ‘Phase-I dose escalation study to evaluate the tolerability and safety of 161 Tb-PSMA in patients with metastatic, castration resistant prostate cancer’. 2024- 2028 Seimbille, Yann Horizon IHI: ‘Accelerate.eu Elevating the future of cancer care with alpha theranostics’. 2024-2029 Seimbille, Yann, and Philip Elsinga NOW-OTP: ‘MOTIVATE: Molecular oncology twins advancing treatment and innovative cancer evaluation’. 2024-2029 Seimbille, Yann, Alexander Vahrmeijer, and Albert Windhorst Dutch Cancer Foundation Grant: ‘First-in-human assessment of a FAP-targeted probe for fluorescence guided surgery of pancreatic cancer’. 2024-2028

Invited Lectures Erik de Blois. ‘Production and quality control of theragnostic radiopharmaceuticals using Alpha’s (IAEA wokshop)’. Centre of Radiochemistry and Nuclear chemistry, Warsaw, Poland. June 2025. Erik de Blois. ‘Production and quality control of Ac-225 labelled radiopharmaceuticals’. NUMERI/NECSA, Pretoria, South Africa. March 2025. Erik de Blois. ‘Ac-225 radiolabelling and GMP’. Institute of nuclear Technology, Online. Nov 2025.

Additional Personnel Amber Piet, MSc – Research assistant Dilan Blijleven - 4th year Chemistry student, Rotterdam University of Applied Sciences. Sept 2024-March 2025. Daily supervisors: Le Li and Hanyue Ma. Pedro Da Silva - 2nd year Chemistry student, Rotterdam University of Applied Sciences. Sept 2025-March 2026. Daily supervisor: Lucie Arnaud.

134

scientific report 2025 | RADIOPHARMACEUTICAL CHEMISTRY

Assistant Professor

Erik de Blois, PhD CLINICAL RADIOCHEMIST: CLINICAL RADIOCHEMISTRY & IMPLEMENTATION Email r.deblois@erasmusmc.nl

Erik de Blois (1981, Zeist) started his MLO training at Zadkine Rotterdam and graduated in 2001 in biotechnology. He then studied part-time at HLO at the HvU Utrecht and graduated in 2005. In the following year he continued his part-time study in pharmaceutical sciences at the University of Utrecht in the direction of “Drug innovation”. He graduated in 2009, after which he started his PhD program part-time at Erasmus MC. In 2014 he successfully defended his thesis: Radiochemical Aspect of receptor Scintigraphy: labeling with radiometals, optimization and radiochemical purity and obtained his doctorate. In the meantime, he started in 2002 as a research analyst in the radiochemistry group at the Nuclear Medicine department at Erasmus MC. After obtaining his master's degree, his position changed to Clinical Radiochemist and for the past 4 years he has also been Head of Quality Control and responsible for the release of radiopharmaceuticals made in-house and started his own research group Clinical Radiochemistry & Implementation (radiochemical aspects of radiopharmacy). Within his position he contributed to the implementation of various (therapeutic) radiopharmaceuticals, including the first registered therapeutic radiopharmaceutical Lu-177 labeled DOTA-TATE (Lutathera).

Clinical implementation of [225Ac]Ac labeled radiopharmaceuticals Targeted alpha therapy with Ac-225 showed to be effective in treating metastatic cancers. However, its complex decay chain requires optimized radiolabeling and quality control conditions. We identified the critical parameters of optimal radiolabeling and established accurate measurement techniques for radiochemical yield and purity for Ac-255 labeled DOTA conjugated radiopharmaceuticals. Therapy with PSMA labeled with actinium-225 can be very effective in patients with metastatic castration resistant PCa (mCRPC). The use of alpha emitting radionuclides will result in more DNA double strand breaks than treatment with beta emitting radionuclides. A limited number of patients have been treated with [225Ac]Ac-PSMA therapy in Germany, and there is still a lack of a good clinical trial. Therefore, a phase I study is necessary to calculate the recommended dose that is both safe and effective. With the treatment of up to 30 patients, we expect to determine the most appropriate cumulative dose for [225Ac]Ac-PSMA treatment. Dose escalation was successfully performed, and more evidence will be collected with a 10 MBq dose of [225Ac] Ac-PSMA.

135

Figure 1. Test and application of 3 different suppliers of Ac225. Metal impurities were measured over time to determine whether the source can be clinically used.

Clinical implementation of [161Tb]Tb-PSMA Therapy with [161Tb]Tb-PSMA is proven to be very effective in patients with metastatic castration resistant PCa in comparison to [177Lu]Lu-PSMA. Radiochemistry is compared to Lu-177 and in vitro and in vivo comparison experiments are recently performed. Additionally, a phase I trial is planned in the near future. Before implementation into cleanroom environment, radiolabeling and stability of the TbPSMA have to be optimized. Reaction conditions and stability were optimized in downscaled form in research labs. The final mixture was formulated and was finally full GMP prepared. For clinical implementation, related equipment, methods and quality controls were calibrated and validated. 3 test runs were performed in the cleanroom and showed that the product was according to all release criteria.

Post-docs

Hanyue Ma, PhD A universal approach to image and treat cancer with novel promising fibroblast activation protein inhibitor Fibroblast activation protein (FAP) is highly expressed in cancer-associated fibroblasts and is abundant in more than 90% of epithelial tumors. FAP is involved in cancer progression, invasion, survival, and treatment resistance, while being barely expressed in normal organs. In recent years, FAP inhibitors based on smallmolecules have been reported as powerful imaging probes for the diagnosis of many types of cancers due to their rapid uptake and impressive tumor-to-background ratios. This further demonstrates the high potential for their use in cancer treatment. We recently developed a series of novel quinolinebased FAP inhibitors (eFAPs) with a broad scope of application in oncology, (i.e., molecular imaging, radionuclide therapy, targeted chemotherapy, imageguided surgery). eFAP analogues showed strong FAP inhibitory activity, high binding affinity, selectivity, and good stability. Both radiolabeled and fluorescent eFAP conjugates showed high and rapid tumor uptake in xenograft tumor models. We believe that our lead candidates can be rapidly translated into clinical trials for imaging and treatment of cancer.

Lucie Arnaud, PhD Development of peptide-based radiopharmaceuticals and personalized neoantigen peptide vaccines Cyclic peptides provide a versatile scaffold for radiotheranostic applications, combining high target affinity, structural stability, and favourable pharmacokinetics. Macrocyclic peptide libraries are being designed using diverse approaches to identify novel candidates with optimized properties for theranostic applications. In parallel, peptide-based radiotracers targeting CXCR4 with dimerization strategy are being developed. CXCR4 is a chemokine receptor overexpressed in multiple cancer types and associated with tumor progression and metastasis, making it a relevant target for both imaging and therapy. In collaboration with the Department of Pulmonary Medicine, personalized neoantigen peptide vaccines are also being developed for pancreatic cancer, within a translational framework bridging preclinical research and clinical application. Our work focuses on drug product development under GMP conditions, including formulation, process development, and quality control.

PhD Students Eline Hooijman, MSc

Dylan Chapeau, MSc

Advisors Erik Verburg, Hugo van der Kuy, Erik de Blois & Stijn Koolen

Advisors Yann Seimbille, Mark Konijnenberg & Erik Verburg

PhD Obtained: 26-11-2025

Synthesis of a trifunctional chelate for multimodality approches

Development of Targeted Alpha Therapy Promising anti-tumor effects were observed in patients treated with small molecules labelled with different types of radioisotopes. α-particle emission provides an advantage over β-emission due to the localized radiation effects and high local cytotoxicity. Subsequently, [225Ac] Ac-PSMA-I&T and [225Ac] Ac-DOTATATE are currently clinically implemented according to full GMP. The first phase I dose-escalation trial with [225Ac]Ac-PSMA-I&T is ongoing with encouraging first results.

Multimodality imaging and/or therapy in a single molecule can be challenging. Therefore, we synthetized new compounds containing a trifunctional linker to combine all modalities to the targeting vector. This synthetic approach avoids the alteration of the biochemical properties of the tracers (e.g., binding affinity). Our strategy relies on a chelate that serves as a bridge between all components of the tracers.

136

scientific report 2025 | RADIOPHARMACEUTICAL CHEMISTRY

Priciana Paraïso, PharmD

Le Li, MSc

Advisors Carolien van Deurzen, Yann Seimbille & Erik Verburg

Advisors Yann Seimbille & Aad van der Lugt

Peptide-based theranostics for breast cancer

Strategies to improve the theranostic potential of fibroblast activation protein (FAP)-targeted radiotracers

Functional noninvasive nuclear imaging (i.e., SPECT, PET) can provide accurate assessment of receptor status in primary and metastatic breast cancer lesions. In this project, we aim at developing a new generation of monomeric and dimeric peptidebased theranostics pointing at two major Achilles’ heels of breast tumors, namely the gastrin releasing peptide receptors (GRPRs) and the human epidermal growth factor (HER2).

The goal is to develop novel radiopharmaceuticals that can be potentially beneficial in the diagnosis and treatment of cancers. A new generation of FAPinhibitors will be designed, synthesized and evaluated in preclinical models to deliver new tracers with improved efficacy and safety profile for imaging and radionuclide therapy.

Carolline Ntihabose, MSc

Lucas de Kruijff, MSc

Advisors Erik Verburg, Harry Hendrikse, Erik de Blois & Stijn Koolen

Advisors Harry Hendrikse, Philip Elsinga, Yann Seimbille & Erik de Blois

The (Pre)Clinical Potential of Targeted Radionuclide Therapy with Novel radionuclides

Development of molecular oncology twins, advancing imaging and treatment of ovarian cancer

Targeted radionuclide therapy (TRT) has revolutionized the treatment of metastasized cancer patients. TRT containing the beta-emitter Lu-177, has shown improved results in cancer treatment. New radionuclides (Ho166, Tb-161, Pb-212, etc.) are necessary to improve current treatments. For example, Tb-161, which has near-identical radiochemical properties. Tb-161 shows pre-clinically more therapeutic effect than Lu-177 TRT, potentially caused by its additional conversion and auger electrons.

Currently, the compound for ovarian cancer (OC) imaging, needed for lesion localization, staging and treatmentresponsive patient selection, is different to the OC therapeutic, which could lead to inadequate treatment. Therefore, the aim is the development of molecular twins: chemically identical compounds containing both nuclides, ensuring identical behaviour in vivo.

137

Sophie Veldhuijzen van Zanten, MD, PhD assistant professor 138

THERANOSTICS FOR CNS AND H&N TUMORS

Top Publications 2025

APPOINTMENT ERASMUS MC DEPARTMENT OF RADIOLOGY & NUCLEAR MEDICINE, RESEARCH APPOINTMENT PRINCESS MÁXIMA CENTER FOR PEDIATRIC ONCOLOGY

Droogers E, IJ Pruis, AW van der Eerden, CE Kromkamp, HL Weber, JA Hardillo, TJ de Jong, AA Harteveld, R Valkema, SE Veldhuijzen van Zanten. Clinical evaluation of [18F]fluoro-2-deoxy-d-glucose positron emission tomography-magnetic resonance imaging ([18F]FDG PET-MRI) and [18F]FDG PET-computed tomography ([18F]FDG PET-CT) for lymph node assessment in head and neck squamous cell carcinoma. Clin Radiol. 2025; 91:107095.

Sophie Veldhuijzen van Zanten holds MSc degrees in Medicine (2011) and Epidemiology (2016). She started her scientific career at the Molecular Neuro-Oncology Laboratory of DanaFarber Cancer Institute and Harvard Medical School, where she developed a strong interest in translational research for patients with difficult-to-diagnose and difficult-to-treat cancers. In 2017, she obtained her PhD in Pediatric NeuroOncology at VU University (Amsterdam). She subsequently completed her combined specialty training in Radiology and Nuclear Medicine at Spaarne Gasthuis (Haarlem, 2016-2019) and Erasmus MC (Rotterdam, 2019-2023), with dedicated sub-specialization in theranostics, neuroradiology, and head and neck imaging. In 2022, she was appointed Assistant Professor at Erasmus MC, where she established a multidisciplinary research program advancing precision imaging and theranostic strategies for patients with difficult-to-treat cancers.

Galldiks N, P Lohmann, M Aboian, RF Jr Barajas, WG Breen, J Ivanidze, DR Johnson, TJ Kaufmann, MM Kim, MJ Mair, G Minnit, M Müther, A Nabavizadeh, JD Palmer, R Rudà, M Smits, N Tolboom, SE Veldhuijzen van Zanten, MC Veronesi, JM Werner, M van den Bent, SM Chang, KJ Langen, I Law, J McConathy, M Mehta, M Preusser, SC Short, R Soffietti, M Vogelbaum, M Weller, PY Wen, JC Tonn, NL Albert. Update to the RANO working group and EANO recommendations for the clinical use of PET imaging in gliomas. Lancet Oncol. 2025; 26:e436-e447. Pruis IJ, V van Dis, SL Maas, RK Balvers, TP van den Bosch, M Segbers, SE Veldhuijzen van Zanten. PSMA radioligand uptake correlates with PSMA expression in high-grade glioma and brain metastases: insights from a prospective PET-MRI guided multiregional biopsy study. Eur J Nucl Med Mol Imaging. 2025; 52:4870-4881.

She serves as principal investigator on multiple competitive national and international research grants and currently supervises five PhD candidates, while three former students have successfully completed their doctoral training at Amsterdam UMC, Princess Máxima Center/UMC Utrecht, and Erasmus MC. In addition to her research activities, she is a national board member of the Dutch Society for Nuclear Medicine (NVNG), a member of the Young Professionals Committee of the European Association of Nuclear Medicine (EANM), and chairs a European registry integrating clinical, imaging, and biological data of children with diffuse midline glioma, driving international collaboration and innovation in precision imaging, theranostics and oncology. s.veldhuijzenvanzanten@erasmusmc.nl

Context

his research line advances precision imaging and theranostics for central nervous system and head and neck cancers, with a focus on diseases that are difficult to diagnose, difficult to reach, and difficult to treat. Theranostics combines molecular imaging with targeted radionuclide therapy (TRT), enabling non-invasive characterization of tumor biology and the delivery of personalized treatment. By visualizing and quantifying target expression throughout the body, theranostic approaches support image-guided diagnosis, staging, treatment selection, and longitudinal response assessment. In addition, they provide unique opportunities to optimize drug delivery, administration routes, dosimetry, and patient selection, thereby accelerating the implementation of precision oncology.

139

Research Projects: Objectives & Achievements Within this research line, we develop and evaluate novel diagnostic and therapeutic radiopharmaceuticals for central nervous system (CNS), head and neck (H&N) and other challenging cancers, spanning molecular target discovery, advanced imaging, and clinical translation.

Primary brain tumours and Brain metastases In a clinical proof-of-concept study led by Ilanah Pruis (former PhD student, current Postdoc), we demonstrated the expression and binding of prostate-specific membrane antigen (PSMA) in patients with adult-type diffuse glioma and brain metastasis. All patients showed uptake of the diagnostic radioligand [68Ga]Ga-PSMA at the tumor site, with minimal to acceptable uptake in healthy brain tissue and other organs. This study provided the first quantified evidence and modelled predictions for the potential of PSMA-based TRT in malignant brain tumors. Additionally, we demonstrated that selective intra-arterial (IA) administration substantially increases tumor uptake compared with intravenous administration. IA delivery of [68Ga]Ga-PSMA in glioma and brain metastasis, and [68Ga] Ga-DOTA-TATE in meningioma, significantly improved tumor-to-liver uptake ratios, qualifying all patients for TRT (Fig. 1). Building on these findings, PhD students Jessica de Jong in collaboration with Jennifer Cheung is leading the clinical translation towards prospective radionuclide therapy trials for glioma. To further optimize treatment efficacy, Mouad Allaoui investigates tumor blood supply and microvascular dynamics using multimodal imaging and quantitative perfusion techniques.

In parallel, Jessica de Jong evaluates the clinical value of O-(2-[18F]-fluoroethyl)-L-tyrosine ([18F]FET) PET-MRI for differentiating progressive disease from treatment-related changes in glioma patients. This multiparametric approach provides complementary metabolic and anatomical information, improving diagnostic accuracy and clinical decision-making. In addition, Sophie Veldhuijzen van Zanten contributed to the updated RANO Working Group and EANO recommendations for the clinical use of PET imaging in glioma, reflecting the Erasmus MC experience with advanced PET-MRI. At the preclinical level, Nina Overdevest (co-supervised by Julie Nonnekens and Justine Perrin) investigates novel molecular targets for glioma theranostics.

Pituitary adenoma For the detection of pituitary neuro-endocrine tumors (PitNETs), formerly known as microadenomas, we introduced a novel diagnostic approach using [18F]FET PETMRI. In the first clinical series of patients with Cushing’s disease and negative or inconclusive MRI findings (n=22, published by Ilanah Pruis), we demonstrated 100% sensitivity and high accuracy for localizing ACTH-secreting PitNETs. This technique shortens diagnostic trajectories, improves surgical and radiation treatment planning, and may enable partial hypophysectomies, thereby reducing the need for lifelong hormone replacement therapy. Consequently, [18F]FET PET-MRI has become the standard of care at Erasmus MC for selected patients with negative or inconclusive MRI findings. The program is currently expanding through the prospective FIND study and a growing network of international collaborations. Within this multicentre initiative, Amber Audhoe evaluates the diagnostic performance of [18F] FET PET-MRI relative to inferior petrosal sinus sampling (IPSS) in patients with Cushing’s disease at Erasmus MC and the University Hospital Basel. In parallel, additional European centers contribute to the validation and broader implementation of this imaging strategy.

H&N tumours and Cancer of Unknown Primary origin

Fig.1. Representative examples of enhanced tumor uptake following selective intra-arterial radioligand administration compared with intravenous delivery.

Esther Droogers evaluated the diagnostic performance of [18F]FDG PET-MRI and [18F]FDG PET-CT for lymph node assessment in patients with head and neck squamous cell carcinoma (HNSCC). Both modalities showed high accuracy in distinguishing benign from malignant lymph nodes, with PET-MRI demonstrating a slight advantage for detecting suspicious lymph nodes. However, the remaining diagnostic uncertainty underlines the need for improved molecular imaging strategies. To address this challenge, our group investigates fibroblast activation protein inhibitor (FAPI)-based imaging,

140

scientific report 2025 | THERANOSTICS FOR CNS AND H&N TUMORS

targeting cancer-associated fibroblasts within the tumor microenvironment. Esther Droogers coordinates the multicenter FAPI for Cancer of Unknown Primary (CUP) study, supported by KWF Kankerbestrijding and developed in collaboration with Missie Tumor Onbekend. The study has progressed rapidly, with the majority of the planned patient cohort enrolled within the first year (Fig. 2).

By combining technological innovation with multidisciplinary clinical translation, this research line aims to establish new paradigms for precision imaging and theranostics and to shape future international standards of care.

Paragangliomas

Veldhuijzen van Zanten, Sophie and consortium partners Medical Delta: ‘Cancer Diagnostics for Sustainable Health Care - Theranostic work package (CARES)’. 2024-2027

In a retrospective cohort study, Esther Droogers (cosupervised by Renske Gahrmann) evaluated the growth and characteristics of extra-adrenal paragangliomas (PGLs) in patients with various Succinate Dehydrogenase (SDHx) mutations using [68Ga]Ga-DOTATATE PET-MRI/CT. The study confirmed the predominantly indolent growth of PGLs in SDHx mutation carriers, with a notably higher risk of tumour progression, malignant transformation, and poor treatment response in patients with SDHB mutations than in those with other SDHx mutations.

Expectations & Directions The rapid expansion of molecular imaging and targeted radionuclide therapies is reshaping precision oncology. By integrating hybrid PET-MRI, advanced quantitative imaging, image-guided drug delivery, and theranostic strategies, this research line aims to improve the diagnosis and treatment of cancers that are difficult to diagnose, difficult to reach, and difficult to treat.

Funding

Veldhuijzen van Zanten, Sophie Dutch Cancer Society: ‘[18F]FAPI PET/CT to Identify Carcinoma of Unknown Primary origin’. 2023-2027 Veldhuijzen van Zanten, Sophie and Julie Nonnekens. Cure Starts Now: ‘Preclinical development of targeted radionuclide therapies for brain tumors (TARGET-FIRST)’. 2023-2027 Veldhuijzen van Zanten, Sophie Erasmus MC Foundation - Daniel den Hoed Fund Young Scientific Talent Award: ‘Initiation of intra-arterial targeted radionuclide therapy for glioma’. 2022-2026 Veldhuijzen van Zanten, Sophie and consortium partners of the European DIPG Network. DIPG Collaborative: ‘European registry for diffuse intrinsic pontine glioma: the SIOPE DIPG Registry’. 2023-2025.

Invited Lectures Sophie Veldhuijzen van Zanten. ‘Quo Vidas Nuclear Medicine’. Plenary Lecture - 38th Annual Congress of the European Association of Nuclear Medicine, Barcelona, Spain. Oct 2025. Sophie Veldhuijzen van Zanten. ‘RANO 2.0 and PET RANO 1.0 in glioma - tools for clinical trials or ready for everyday practice?’ 20th Meeting of the European Association of Neuro-Oncology (EANO), collaborative session with the Chinese Society of Neuro-Oncology (CSNO), Prague, Czech Republic. Oct 2025. Sophie Veldhuijzen van Zanten. ‘Theranostics for CNS and H&N tumors’. 4th Conference of the 180°N Norwegian Nuclear Medicine Consortium, Oslo, Norway. March 2025.

Fig.2. [18F]FAPI PET demonstrates intense uptake in a pancreatic lesion that is inconspicuous on [18F]FDG PET, illustrating the complementary diagnostic value of FAPI imaging.

Current efforts focus on the clinical implementation of novel imaging biomarkers and targeted radionuclide therapies, supported by prospective clinical studies and international collaborations. At the same time, innovations in intra-arterial radioligand delivery and quantitative dosimetry are enabling personalized treatment strategies for diseases that have remained beyond the reach of conventional approaches.

Additional Personal Mouad Allaoui, MSc Biomedical Technology and Physics, research assistant Department of Radiology & Nuclear Medicine, Erasmus Medical Center, Rotterdam. Ongoing projects “Quantitative DSA validation using perfusion MRI”, “Semi-automated segmentation of FAPI PET”, “Semi-automated segmentation of FET PET”, supervised by Ilanah Pruis (daily supervisor), Esther Droogers (daily supervisor), Theo van Walsum, Frank te Nijenhuis and Sophie Veldhuijzen van Zanten.

141

Romy van der Groef, 3rd-year PhD student Department of Internal Medicine / Endocrinology, Erasmus Medical Center, Rotterdam. Ongoing project “Pasireotide Induces Long-term Cystic Degeneration of Somatotrophic Pituitary Neuroendocrine Tumors (PitNETs)”, supervised by Sophie Veldhuijzen van Zanten, Julie Refardt and Sebastian Neggers. Da Hao, MSc Pharmacology, Leiden University Medical Center (LUMC), in collaboration with Amsterdam UMC (location VUmc). Ongoing project “Physiologically-Based Pharmacokinetic Modeling of [68Ga]Ga-PSMA-11 in Primary and Secondary Brain Tumors to Facilitate PSMAtargeted Theranostics”, supervised by Ilanah Pruis (daily supervisor) and Sophie Veldhuijzen van Zanten. Mimi Welte, BSc Erasmus University College, Major Molecular and Cellular Biology, 11/2024 - 06/2025 “Immune modulation by Ionizing Radiation in Pediatric CNS tumors”, supervised by Nina Overdevest (daily supervisor).

Post-doc

Ilanah Pruis, PhD Theranostic PET-MRI for Brain Tumours

Despite advancements in precision cancer medicine, patients with brain tumours still face a one-size-fits all approach. Patient outcomes are dismal and have not improved for decades, indicating the significant need for innovation. I aim to integrate advanced hybrid imaging with targeted radionuclide-strategies for precise diagnosis and therapy, referred to as theranostics, in neuro-oncology. The potential of theranostics for brain tumour management has recently gained significant interest. My research bridges clinical molecular imaging with the underlying pathophysiology of brain tumours, prioritizing (radio)biological validation and clinical implementation focused on delivering personalized, hybrid image-based approaches for patients with a brain tumour. My expertise lies in implementing hybrid PET-MRI in neuro-oncology. Key projects of my research include: 1. Initiating molecular imaging projects to validate novel theranostic targets in brain tumours; 2. Launching translational clinical studies to assess the diagnostic and therapeutic potential of preclinically validated tracers such as PSMA-based tracers for glioma and brain metastasis and FET-based tracers for pituitary adenomas (PitNETs); 3. Proof-of-concept investigations into advanced delivery techniques such as super-selective intra-arterial delivery.

142

scientific report 2025 | THERANOSTICS FOR CNS AND H&N TUMORS

PhD Students Jessica de Jong, MD, MSc

Esther Droogers, MD, MSc

Advisors Erik Verburg, Marion Smits, Sophie Veldhuijzen van Zanten & Marjolein Geurts

Advisors Sophie Veldhuijzen van Zanten, Debbie Robbrecht & Erik Verburg

Hybrid Imaging and Targeted Radionuclide Therapy for Glioma

[18F]-FAPI PET-CT in patients with carcinoma of unknown primary origin

My research focuses on the use of radiopharmaceuticals for both diagnostic and therapeutic purposes, i.e. theranostics, in the management of glioma. Highlights of my work Include the clinical utility of hybrid [¹⁸F]FET PET‑MRI to distinguish progression from treatment effects and, an upcoming phase 1 trial delivering intra‑arterial [¹⁷⁷Lu]Lu‑PSMA‑I&T as salvage therapy for recurrent high‑grade glioma.

This project focusses on imaging bioprosthetic valves with CT and MRI. In particular patients that underwent a percutaneous pulmonary valve implantation (PPVI). CT seems a reliable tool for risk assessment of coronary artery compression. There seems no relevant change in RVOT to coronary distance and coronary lumen diameter after PPVI. Conduit expansion does not seem to affect the relationship between the pulmonary trunk and coronary arteries after implantation. In a second part of the project we are now evaluating flow patterns after PPVI with 2D and 4D flow MRI.

Jennifer Cheung, MSc Advisors Sophie Veldhuijzen van Zanten, Stijn Koolen & Harry Hendrikse

Amber Audhoe, MSc Advisors Julie Refardt, Sophie Veldhuijzen van Zanten & Sebastian Neggers

Brain tumor imaging with radiopharmaceuticals My PhD focuses on innovative treatment strategies for brain tumors and improving drug delivery. We are setting up a clinical trial on intra‑arterial administration of an alpha‑emitter in meningioma, enabling localized therapy with minimal systemic toxicity. We also aim to investigate the pharmacokinetics and bioavailability of C‑14‑labeled drugs in glioblastoma.

Use of [18F]FET PET-MRI to improve detection of pituitary adenomas in Cushing’s disease My prospective study compares the diagnostic accuracy of bilateral inferior sinus petrosal sampling (IPSS) to [18F]FET PET-MRI in Cushing’s Disease. Because MRI often shows microadenomas/inconclusive findings, IPSS is required to confirm pituitary ACTH hypersecretion, but it is invasive and does not localize the adenoma precisely, highlighting the need for improvement.

Nina Overdevest, MSc Advisors Julie Nonnekens, Sophie Veldhuijzen van Zanten & Justine Perrin

Development and optimization of targeted radiopharmaceutical therapies for pediatric brain tumors; a world-first translational study Currently, there is a lack of effective treatments available for pediatric medulloblastoma and diffuse intrinsic pontine glioma/diffuse midline glioma (DIPG/DMG). Therefore, the goal of this project is to determine through preclinical research whether targeted radionuclide therapy (TRT) is a potent treatment option for medulloblastoma and DIPG/ DGM.

143

144

Focus Area

CLINICAL IMAGING

The aim is to investigate the clinical value of (new) imaging technologies and imaging biomarkers, following a structured order of investigations. The robustness of imaging biomarker extraction is assessed with a focus on accuracy, repeatability and reproducibility. The diagnostic accuracy of new imaging technology is investigated by comparing this imaging data to reference standards. The clinical relevance is assessed by evaluating diagnostic confidence regarding clinical decision-making and impact on treatment planning. The prediction of outcome or treatment response based on quantitative imaging biomarkers, radiomics features and deep-learning algorithms is evaluated. Multicenter (clinical) trials are used to assess and evaluate imaging biomarkers of disease activity and response to treatment.

145

Aad van der Lugt, MD, PhD full professor 146

IMAGING IN NEUROVASCULAR DISEASE

Top Publications 2025

Aad van der Lugt graduated from Erasmus Medical School in 1988. Prior to specializing in neuroradiology at Erasmus Medical Center, he worked as a junior doctor in Surgery and Intensive Care in Tilburg and Eindhoven. He obtained his PhD from Erasmus University Rotterdam in 1996 with a dissertation entitled “Intravascular Ultrasound: Validation and Clinical Application.”

Van der Sluijs PM, R Su, S Cornelissen, AC van Es, A Lycklama, GJ Nijeholt, PJ van Doormaal, WH van Zwam, DW Dippel, T van Walsum, A van der Lugt. Assessment of automated TICI scoring during endovascular treatment in patients with an ischemic stroke. J Neurointerv Surg. 2025; 17:830836.

Over the years, he has played a leading role in neuroradiological research. He chaired the Neuroradiology research program from 2002 to 2015 and was appointed Professor of Neuroradiology and Head & Neck Radiology in 2010. Since 2022, he has served as Chair of the Department of Radiology & Nuclear Medicine at Erasmus MC. He represents the Netherlands on the Euro-BioImaging Board.

Schoon BA, D Hansen, B Roozenbeek, J Oude Groeniger, W van der Steen, A van der Lugt, M Kappelhof, YB Roos, CB Majoie, F van Lenthe, DW Dippel. Neighborhood Socioeconomic Status and the Functional Outcome of Patients Treated With Endovascular Thrombectomy for Ischemic Stroke. Neurology 2025; 105:e213615.

His research focuses primarily on imaging and treatment of acute stroke. In addition, he leads national research efforts within the Dutch acute stroke CONTRAST Consortium.

Van Hulst PL, RM van de Wijdeven, E Venema, FM Pinckaers, MG Hunink, A van der Lugt, DW Dippel, HF Lingsma, D Bos, B Roozenbeek. A Decision-Analytic Model to Evaluate Cost-Effectiveness of Regional Implementation of a Mobile Stroke Unit. Neurology 2025; 105:e213834.

a.vanderlugt@erasmusmc.nl

Context

his research program is focused on the role of imaging biomarkers in neurovascular diseases with a strong emphasis on ischemic stroke. Imaging in acute stroke aims to support in the diagnosis, assessment of the severity and reversibility of ischemia. Patient selection for treatments like intravenous thrombolysis or endovascular thrombectomy relies on imaging biomarkers. These markers guide decisions by identifying eligible patients. In the sub-acute phase, imaging can assist in evaluating the underlying cause of ischemic stroke. Recently, there’s been increased attention to atherosclerosis beyond the carotid bifurcation. Optimal secondary preventive measures can be better tailored by subdividing patients based on their presumed etiology of ischemic stroke. This approach aims to enhance patient outcomes. This research program encompasses technical development, evaluation of image analysis algorithms, and clinical validation of imaging biomarkers.

147

Research Projects: Objectives & Achievements Endovascular treatment in patients with acute stroke: beyond the MR CLEAN studies Treatment with intravenous (IV) alteplase, aimed at early reperfusion, has proven effective for patients with acute ischemic stroke. Approximately 25% of patients experiencing acute anterior circulation ischemic stroke exhibit symptoms caused by proximal occlusion of a major intracranial artery. Endovascular treatment (EVT) significantly increases the likelihood of recanalization in patients with acute ischemic stroke resulting from proximal intracranial arterial occlusion. The MR CLEAN Study demonstrated that intra-arterial treatment, administered within 6 hours after stroke onset, is both effective and safe. Since our landmark paper, multiple studies have further highlighted the beneficial aspects of EVT. Within the HERMES Collaboration, individual patient data have been leveraged to investigate additional research questions. These inquiries have focussed on optimal patient selection for treatment and improvement of overall outcomes. In 2017 the CONTRAST-consortium (www.contrastcon-sortium.nl) was established creating a biobank infrastructure for stroke research in the Netherlands. This infrastructure included data, blood samples, thrombus and imaging. Three new trials have been conducted 1) to evaluate the effect of peri-procedural medication (MR CLEAN-MED), 2) to evaluate the effects in patients presenting between 6 and 12 hours after the event (MR CLEAN-LATE) and 3) to evaluate the benefits of direct IAT without prior IVT (MR CLEAN-NoIV). All trials have been completed and their results have been published. In 2023 the Consortium was extended by 5 years and secured new funding. The CONTRAST 2.0 Consortium aims to enhance the chances of a good recovery for a large group of patients with cerebral infarction, brain hemorrhage, or subarachnoid hemorrhage. This will be achieved by developing new treatments in the laboratory, improving treatment logistics, and initiating six large clinical trials. Each study will test one or more new treatments or answer important research questions. All studies will be benefit from a central core lab assessment of CT and MRI scans of the brain along with a central database and standardized statistical analysis.

In general, stroke thrombi consist of four main components: red blood cells, fibrin, platelets and white blood cells. In a MR CLEAN Registry biobank study, we have shown that the amount of red blood cells and fibrin/ platelets are related to the origin of the thrombus: red blood cell-rich clots were more likely to originate from large artery atherosclerosis, while fibrin/platelet-rich clots were more likely to have a cardioembolic source. CT imaging at admission allows us to assess various thrombus characteristics, including attenuation, length, distance from the internal carotid artery terminus, and permeability (or “perviousness”). We have demonstrated that thrombus CT features correlate with its composition (e.g., red blood cells and fibrin/platelets). Importantly, these CT characteristics also align with the thrombus’s origin, validating our hypothesis based on histological studies. This is crucial, as histological analyses can be time-consuming, costly, and subject to bias. In collaboration with the biomechanical department at Technical University Delft, we conducted the firstever study to mechanically characterize human stroke thrombi. Our findings confirmed that fibrin/platelet-rich thrombi are mechanically stiffer. A highly debated topic in the field of EVT revolves around device selection: stent retrievers (SR) have long been the standard, but contact aspiration (CA) has rapidly gaining popularity. Researchers and clinicians speculate that thrombus type could serve as a guide for choosing the optimal first-line EVT device. Specifically, they believe that fibrin/platelet-rich thrombi, which are stiff and resistant to thrombectomy, might benefit more from CA. However, a MR CLEAN Registry substudy revealed a different outcome. Contrary to expectations, we found that this correlation did not hold true.

Role of Thrombus With EVT extracted thrombi have become available for histopathologic analysis providing a unique opportunity to study several critical relationships. We aimed to 1) explore how imaging features of the thrombus relate to its composition; 2) Investigate how thrombus composition influences biomechanical properties and 3) understand the impact of thrombus composition on successful recanalization.

Figure 1. Clot Computed Tomography (CT) characteristics. (i) Representative red blood cell (RBC) group CT images for a noncontrast (NCCT) and contrast-enhanced (CECT) scan (all 90 x 103 platelets/μl). (ii) Scatterplots of RBC content versus NCCT density and CECT density increase. Points are coloured according to platelet content.

148

scientific report 2025 | IMAGING IN NEUROVASCULAR DISEASE

Project • EVT for acute ischemic stroke: lessons learned from the occluding thrombus (Nikki Boodt)

Socioeconomic status Socio-economic status (SES) has previously been identified as affecting both stroke incidence and functional outcome after stroke. Although the disparity in stroke morbidity and mortality between low- and higher-income countries is well-known, there is evidence that patients with relatively low SES in high-income countries are also disproportionately affected. Few studies have focused on patients treated with endovascular thrombectomy (EVT). we are interested in the association between the neighborhood socio-economic status (nSES) and outcome after thrombectomy. Project • Socioeconomic status and outcome of endovascular treatment (Bridget Schoon)

Invited Lectures Aad van der Lugt. ‘Blood-based Biomarkers, liquid biopsy, AI and the role of Imaging in follow-up’. International Society of Strategic Studies in Radiology, Dublin, Ireland. Aug 2025. Aad van der Lugt. ‘Nuclear Medicine & Radio-pharmacy and the integration with anatomical imaging’. Erasmus MC, Nuclear Medicine symposium, Rotterdam, the Netherlands. March 2025. Aad van der Lugt. ‘AI in Cancer Imaging, A mile-high overview’. International Cancer Imaging Society, The Hague, the Netherlands. Sept 2025. Aad van der Lugt. ‘State -of-the-art CT solutions for stroke diagnosis & treatment’. Symposium The new role of CT in Stroke diagnosis & treatment, European Congress of Radiology, Vienna, Austria. Feb 2025.

Expectations & Directions The expected results of CONTRAST 2.0 are: 1) Faster and better treatment healthcare organization.

through

improved

2) Expanding the possibilities for treating cerebral infarction with a catheter and medication. 3) Greater chance of success in treating cerebral infarctions and subarachnoid haemorrhages with a catheter. 4) New effective treatments for brain haemorrhage. 5) Better advice on how to guide patients to further recovery and better functioning after a stroke. 6) Prospects for new treatments in the future through animal experiments with more knowledge about the development of cerebral infarctions and ways to slow it down with medication.

Funding Van der Lugt, Aad, Yvo Roos, and partners: IMPULSE program Dutch Heart Foundation, Brain Foundation Netherlands in collaboration with the Dutch Cardio-Vascular Alliance: ‘CONTRAST2.0, consortium for new treatments for acute stroke’. 2023-2028 Van der Lugt, Aad, Yvo Roos, and CONTRAST Consortium Partners: Medtronic: ‘CONTRAST2.0, consortium for new treatments for acute stroke’. 2024-2027

149

Post-doc

Rob van de Graaf, PhD Evaluation of microcirculatory disfunction following stroke

Every year, over 1500 patients in the Netherlands presenting with an occlusion of a major supplying artery receive EVT with or without prior thrombolysis. Despite the substantial beneficial effect of current EVT on patient functional outcomes, still about 54% of patients do not recover to functional independence and die or are severely disabled during life. Currently, procedural success of EVT strategies is scored by visual evaluation using digital subtraction angiography (DSA), which displays the vessel status at the clot location and the distal vessel bed. Although macrovascular distal vessel bed reperfusion may intuitively be linked to reperfusion in the microvasculature the underlying pathophysiology of reperfusion in the microcirculation is different and currently not evaluated. In my work I aim to evaluate the intracranial microcirculation in the setting of stroke. One of the projects I am working on is whether an established technique from the field of interventional cardiology to measure the absolute flow and vessel bed resistance can be applied to the neurovasculature. Using a simple and accurate (operator independent) technique of thermodilution technique in combination with a dedicated temperature and pressurewire, both flow and vesselbed resistance can be measured. Ultimately these measurements will guide treatment decisions in the setting of both ischemic and hemorrhagic stroke.

PhD Students Nikki Boodt, MD, MSc

Bridget Schoon, MD, MSc

Advisors Aad van der Lugt, Diederik Dippel & Hester Lingsma

Advisors Aad van der Lugt, Diederik Dippel & Bob Roozenbeek

PhD Obtained 27-05-2025

Socio-economic status and outcome of endovascular treatment

EVT for acute ischemic stroke: Lessons learned from the occluding thrombus

Although the disparity in stroke morbidity and mortality between low- and higher-income countries is wellknown, there is evidence that patients with relatively low SES in high-income countries are also disproportionately affected. Mu research aims to investigate the association between the socio-economic status and outcome after mechanical thrombectomy.

My research aims to evaluate the role of the occluding thrombus in achieving successful reperfusion. I focus on the histological, mechanical and imaging characteristics of intracranial large vessel thrombus, and their relationship with procedural and clinical outcomes, to help improve EVT efficacy.

150

scientific report 2025 | IMAGING IN NEUROVASCULAR DISEASE

151

Marion Smits, MD, PhD full professor 152

NEURORADIOLOGY

Top Publications 2025

Marion Smits is an internationally active Neuroradiologist who combines clinical work with scientific research as full professor of Neuroradiology. She is also Medical Delta Professor and holds an appointment as full professor of Neuroradiology at the TU Delft. Marion studied Medicine at Maastricht University and worked as a junior doctor in the United Kingdom before specializing as a Neuroradiologist at Erasmus MC. She obtained her PhD cum laude from Erasmus University Rotterdam in 2008. Marion is chair of Research on the board of the Radiological Society of the Netherlands, past-chair of the Brain Tumor Group Imaging Committee of the European Organization for Research and Treatment in Cancer, and active in key national and international organizations. She also served on the ESR Executive Council as chair of the Research Committee until March 2025.

Cakmak M, S Mohammadian, VC Keil, JW Schouten, PC De Witt Hamer, T Van der Vaart, RK Balvers, IJ Wamelink, F Barkhof, MJ Van den Bent, M Vries, M Smits. How useful is contrast-enhanced MRI in the long-term surveillance of glioma? A multicentre retrospective longitudinal cohort study. Eur Radiol 2025; 35:4257-4266. Topff L, L Petrychenko, N Jain, S Lingier, J Bertels, P Astudillo, M Prosec, P Menéndez FernándezMiranda, O Gevaert, M Smits, S Derks, E Verhaak, P Hanssens, E Marco de Lucas, R Sutil, P Dominguez Echávarri, A Negoita, E Visser, D Corral Fontecha, L Braun, D Brandsma, J Visser, E Ranschaert, K Groot Lipman, R Beets-Tan. A Data-Centric Approach to Deep Learning for Brain Metastasis Analysis on MRI. Radiology 2025; 315:e242416.

marion.smits@erasmusmc.nl

Vallentgoed WR, Y Hoogstrate, KA Van Garderen, L Van Hijfte, E Van Dijk, MC Kouwenhoven, JM Niers, K Draaisma, I Martin, WW De Leng, CM Tesileanu, I De Heer, M Diepeveen, A Lavrova, PP Eijk, M Bühler, W Wick, PM Clement, M Sanson, E Franceschi, T Gorlia, V Golfinopoulos, M Weller, T Weiss, PA Robe, JM Kros, M Smits, M Van de Wiel, B Ylstra, RG Verhaak, MJ Van den Bent, BA Westerman, P Wesseling, PJ French. Evolutionary trajectories of IDH-mutant astrocytoma identify molecular grading markers related to cell cycling. Nat Cancer 2025; 6:16931713.

Context

his research line is focused on the human brain’s function and (micro)structure under physiological and particularly under pathological conditions. Physiological and functional MR neuroimaging techniques combined with artificial intelligence image analysis are uniquely suited to study the human brain in vivo. These techniques include functional MRI (fMRI), diffusion, metabolic and perfusion imaging. The clinical applicability of these various imaging techniques and their findings are an important aspect of this research line. The research is performed in a continuous interplay between fundamental imaging research and clinical practice, with a primary focus on Neuro-Oncology. This means that there is a close collaboration with clinically as well as technically oriented researchers, in particular within the Erasmus MC Brain Tumor Center and the Medical Delta. To ensure true clinical value and implementation of this research, there is also a strong link with the Erasmus School of Health Policy & Management for health technology assessment in the context of the Convergence.

153

Research Projects: Objectives & Achievements Clinical validation Advanced MR neuroimaging techniques are developed together with the MRI Physics group led by Prof. Juan Hernandez Tamames (page 34). Further technical development is achieved through collaboration within the Medical Delta, in particular with the MRI physics group at Leiden University Medical Center (LUMC, Prof. M.P.J. van Osch) and Imaging Physics at Delft University of Technology (TU Delft, Dr. F.M. Vos). Such techniques are explored for their potential to provide imaging markers of disease, within the research line led by Dr. Esther Warnert (page 158). Here, we focus on the transition from these (technical) labs to clinical practice. One such novel techniques is chemical exchange saturation transfer (CEST). CEST is still mostly in the research stage. Together with King’s College London (Prof. G. Barker, Dr. T. Wood) we implement amide proton transfer (APT)-CEST in clinical practice and imaging genomics studies. Quantitative physiological MR imaging of the brain is of great interest both for research and clinical applications and is becoming a realistic possibility with the availability of several imaging sequences with (clinically) acceptable scanning times. Quantitative measures of brain structure and function allow for reference values to distinguish normal from abnormal conditions, follow-up studies sensitive to subtle changes over time, and the exchange or pooling of data across centers. Perfusion imaging is a commonly used technique for tumor characterization and surveillance, with varying observations for the different techniques (Anna Lavrova). Other techniques that are being developed and assessed in a clinical environment are more advanced vascular and metabolic imaging markers (Laura Kemper and Jessica de Jong), tissue relaxation measurements (Noemi Sgambelluri), and MR fingerprinting approaches (Karen van der Werff). The clinical implementation and value of existing imaging techniques also require scientific scrutiny, especially where there is heterogeneity and variation throughout hospitals. For instance, there is a wide variety in the use of perfusion MRI for brain tumor diagnosis and follow-up, even within The Netherlands. In the PERISCOPE project we collaborate with Dr. Brenda Leeneman and Dr. Seamus Kent (Erasmus School of Health Policy & Management) to assess the value of perfusion MRI from a clinical and cost-effectiveness perspective in a large multicenter observational study.

Diagnostics: Virtual biopsy For diagnosis, imaging genomics of brain tumors has gained substantial relevance with the recently updated classification of brain tumors by the World Health

Organization which relies heavily on tumor genetics. The non-invasive assessment of tumor genotypes is important for treatment decisions and follow-up. We develop and use physiological MR imaging and advanced postprocessing techniques for imaging genomics of adulttype diffuse glioma in a multicenter setting, to obtain a so-called virtual biopsy (Juancito van Leeuwen, Bas Dille, Gonzalo Mosquera Rojas, Tareq Abdel Alim) with advanced image analysis techniques developed by the Biomedical Imaging Group Rotterdam led by Prof. S. Klein (64). This multidisciplinary project is performed in close collaboration with the departments of Neurology (Prof. M.J. van den Bent, Dr. M. Geurts, Dr. M. Wijnenga, Prof. P. French), Pathology (Dr. S.N. Maas), Neurosurgery (Prof. A.J.P.E. Vincent, Dr. E.M. Bos), as well as the Radboudumc Nijmegen (Prof. G. Litjens, Dr. D. Henssen, Dr. B. Kusters). To study the interaction with and acceptance by various stakeholders of such diagnostic artificial analysis tools, we also work together with the Erasmus School of Health Policy & Management (dr. B. Leeneman, Ikenna Asuzu). For real-world validation we work together with the National Hospital for Neurology and Neurosurgery of University College London (Dr. M. Grech Sollars, Prof. T. Yousry) and Nottingham University Hospitals NHS Trust (Dr. S. Thust).

Surrogate markers & surveillance strategies Especially in the context of newly developed treatments, accurate diagnosis and response assessment is of the utmost importance. Through collaboration with the European Organization for Research and Treatment of Cancer (EORTC) imaging markers of outcome after treatment are investigated. This now also includes an investigation into the role of advanced MRI and PET imaging. MR imaging-based assessment of brain tumors is traditionally heavily reliant on contrast-enhanced acquisition. Especially in patients with long survival times, this results in large cumulative contrast-agent exposure. Concerns with Gadolinium-based contrast agents are increasing, not only in terms of patient safety, but also due to their burden on the environment. Together with Amsterdam UMC (Dr. V.C. Keil, Prof. F. Barkhof, Prof. P. de Witt Hamer, Dr. M. Vries) we investigate alternative strategies of MRI-based surveillance in long-term survivors after glioma diagnosis. Additionally, we investigate whether contrast-agent administration can be obviated in patients with meningioma in collaboration with Neurosurgery (Dr. E. Bos).

Image guidance for invasive tumor treatment Brain tumor treatment through surgery or targeted radiation therapy aims to balance maximal reduction of tumor burden versus minimal damage to eloquent brain structures. Thus, both precise tumor delineation and reliable identification of functionally important tissue at risk is required pretreatment. Several imaging techniques are developed and evaluated for this purpose. With diffusion MRI white matter tracts at risk can be identified prior to surgery using MRI-based tractography. With the

154

scientific report 2025 | NEURORADIOLOGY

TU Eindhoven (prof. L. Florack) and Elisabeth Tweesteden Ziekenhuis in Tilburg (prof. G.J. Rutten) we develop advanced tractography tools that provide robust results in a clinical setting (Mégan van de Veerdonk). With functional MRI (fMRI) we aim to gain insight in motor, language, and memory processing both under physiological and pathological conditions. Our collaboration with the functional Ultrasound group within CUBE (Dr. P. Kruizinga, Dr. S. Soloukey) opens opportunities to correlate intra-operative findings with preoperative fMRI in terms of functional imaging characteristics and validity. Additionally, tumor vascularization can be assessed in great detail with so-called microDoppler ultrasound, providing important information for MRI based assessment of tumor vascularization (Ahmad Alafandi). Optimization of the tumor target volume using advanced MRI is the aim of the project entitled ‘Hitting the mark’ (Patrick Tang) in a close collaboration with Radiotherapy (Prof. R. Nout, Dr. A. Mendez Romero).

Implementation Through leading roles in European organizations, the clinical aspects of imaging protocols and protocol harmonization are assessed and disseminated. Despite the fact that many physiological MR neuroimaging techniques have already been available for decades, some even being extensively used for fundamental research, their application in patient studies is still relatively limited. Clinical implementation is even less frequent, due to the fact that imaging findings from group studies commonly fail to make the essential transition to the individual patient level. It is the ultimate aim of this research line to provide imaging markers of brain physiology and disease that are directly clinically applicable and widely accessible.

Expectations & Directions My additional appointments as Medical Delta Professor and full professor of Neuroradiology at TU Delft provide strong avenues for multicenter, multidisciplinary collaboration in health-tech development. These multidisciplinary collaborations provide the context for advancing and expanding our current studies, in a continuous interplay between physiological imaging research and highly expert clinical practice. With active positions in and connections with the EORTC brain tumor and imaging groups, the ESR, the European Association for NeuroOncology (EANO), and the ESMRMB working group on Glioma MR imaging (glimr. eu), future efforts are directed at furthering the role of physiological neuroimaging in clinical research and practice on an international level.

Funding De Geus-Oei, Lioe-Fee, Jeroen Kalkman, Sophie Veldhuijzen van Zanten, Myriam Menzel, Thijs van Osch, and Marion Smits Medical Delta 3.0: ‘Cancer Diagnostics for Sustainable Health Care (CARES)’. 2024-2028

Hernandez-Tamames, Juan, Marion Smits, Dirk Poot, and Laura Nunez Gonzalez TKI-LSH: ‘Gadolinium-Free Enhancement with Magnetic Resonance Imaging Synthesis – GEM’. 2024-2028 Florack, Luc, Geert-Jan Rutten, and Marion Smits NWO KIC: ‘Bringing tractography into daily neurosurgical practice.’ 2023-2028 Smits Marion VICI: ‘Virtual biopsy: paving the way towards reality.’ 2023-2028 Smits, Marion, Thijs van Osch, Dirk Poot, Stefan Klein, and Juan Hernandez Tamames NWO-TTW Open Technology Program: ‘Vascular Signature Mapping of Brain Tumor Genotypes.’ 2019-2025

Invited Lectures Renske Gahrmann. ‘Brain Tumours: Guiding Therapy with PET’. ECR, Vienna, Austria. Feb 2025. Marion Smits. ‘Setting up task-based fMRI.’ ESNR-ESMRMB European Course on Advanced Imaging Techniques in Neuroradiology, Milan, Italy. Nov 2025. Marion Smits. ‘Hot topic round table: Are we redundant? The future of radiology professions.’ ESMRMB annual meeting, Marseille, France. Oct 2025. Marion Smits. ‘Closing Keynote lecture - Reducing contrast agent use for neuro-oncology.’ GREC pre-congress ESMRMB annual meeting, Marseille, France. Oct 2025. Marion Smits. ‘Advances in neuro-oncological imaging.’ International Cancer Imaging Society, The Hague, the Netherlands. Sept 2025. Marion Smits. ‘Technical advances in neuro-oncology: what is the future: radiomics or better sequences?’ ESNR annual meeting, Istanbul, Turkey. Sept 2025. Marion Smits. ‘Histopathological and molecular correlates of glioma imaging.’ European Association of Neurology, Helsinki, Finland. June 2025. Marion Smits. ‘Imaging response of brain tumours.’ ESOIEORTC workshop Imaging in assessing response to cancer therapy, Brussels, Belgium. May 2025. Marion Smits. ‘Keynote lecture: Genetic and molecular biomarkers - radiological impact.’ European Course of Neuroradiology, Prague, Czech Republic. May 2025. Marion Smits. ‘Post-treatment glioma imaging: strategy; findings; pitfalls.’ European Course of Neuroradiology, Prague, Czech Republic. May 2025. Marion Smits. ‘Brain tumour imaging after treatment.’ ANZSNR annual scientific meeting, Perth, Australia. March 2025.

155

Marion Smits. ‘Brain tumour imaging: from phenotype to genotype, and everything in between.’ ANZSNR annual scientific meeting, Perth, Australia. March 2025.

S. Stanakzai – Medicine MSc student (Erasmus MC)

Marion Smits. ‘The pathophysiology of imaging findings in ARIA-E and ARIA-H - a pictorial overview.’ ANZSNR annual scientific meeting, Perth, Australia. March 2025.

F. Caglayan – Medicine MSc student (Erasmus MC)

L. ter Wisscha – Technical Medicine BSc student (TU Delft)

L. Wansink – Technical Medicine BSc student (TU Delft)

Marion Smits. ‘What’s new in perfusion MRI.’ ANZSNR annual scientific meeting, Perth, Australia. March 2025. Marion Smits. ‘How quantitative imaging and biomarkers are changing neuroradiology.’ ANZSNR annual scientific meeting, Perth, Australia. March 2025. Marion Smits. ‘Clinical perspectives on perfusion MRI for brain tumour assessment & follow-up.’ ISMRM perfusion workshop, Pamplona, Spain. March 2025.

PhD Students

Marion Smits. ‘Imaging of intracranial masses.’ ECR 2025 annual meeting, Vienna, Austria. Feb 2025.

Mégan van de Veerdonk, MD, MSc

Additional Personnel

Advisors Marion Smits, Geert-Jan Rutten & Bart Brouwers

M. Mannil - PhD student (UMCN St. Radboud) R.W. Sol-Kloet – PhD student (UMCG)

Bringing tractography into daily neurosurgical practice Tractography is still rarely used in clinical neurosurgical routine due to a variability in technical and conceptual analyses of diffusion MRI and in functional-anatomical definitions of white matter tracts. In this research project, we aim to create a standardized turn-key system to improve daily neurosurgical workflow for functional and oncological neurosurgery.

C. Tseng - PhD student (TU Delft) D. van Dorth - PhD student (LUMC) B. Monkel - Medicine MSc student (Erasmus MC)

Post-docs

Renske Gahrmann, MD, PhD Hybrid imaging of the Head and Neck Hybrid PET MRI provides simultaneous metabolic and high-resolution anatomical information and offers a platform for more objective evaluation of complex diseases. In head and neck cancer, assessment of primary tumors and cervical lymph nodes remains inconsistent, while the positive predictive value of FDG PET for nodal metastases is limited. Using a large retrospective PET MRI cohort, my project in collaboration with Gyula Kotek and Anita Harteveld develops motion corrected, multimodally registered datasets to train automated segmentation and malignancy prediction models, enabling standardized delineation and quantitative biomarker extraction. In parallel, a prospective pilot study investigates 18F FAPI PET/MRI in Graves’ orbitopathy (GO), in collaboration with the Rotterdam Eye Hospital, where distinguishing active inflammation from fibrotic, burnt out disease is clinically challenging. By comparing FAPI uptake in active GO, inactive GO, and non GO controls, the study evaluates whether fibroblast targeted imaging can serve as an objective marker of disease activity. Together, these projects demonstrate how advanced hybrid imaging and computational analysis can improve diagnostic consistency and support clinical decision-making across oncology and inflammatory disorders.

156

scientific report 2025 | NEURORADIOLOGY

Tareq Abdel Alim, PhD Virtual Biopsy: Towards non-invasive brain tumor characterization I work on the Glioma Virtual Biopsy project, where I develop integrated radiopathomics models to study relationships between histopathological markers and MRI features in glioma, with the aim of enabling non-invasive tumor characterization from MRI alone. Using novel interpretability methods and computationally efficient deep learning approaches, this research aims to significantly reduce patient burden and risk by decreasing the need for surgical biopsies, while also supporting more accessible and equitable care. This may benefit settings with limited access to expensive molecular diagnostics or neurosurgical facilities. During my PhD, I specialized in multimodal image analysis to understand how congenital dysmorphologies (and their treatments) impact the craniofacial shape and development, with an emphasis on non-invasive imaging techniques. Through the supervision of PhD students and in my role as chair of the Medical Imaging Working Group within the European Reference Network (ERN-CRANIO), focused on equitable access to specialized care for children with congenital craniofacial conditions, I remain actively involved in this research line.

Bas Dille, MSc

Ahmad Alafandi, MD, MSc

Advisors Marion Smits, Stefan Klein, Niek Maas & Geert Litjens

Advisors Marion Smits & Pieter Kruizinga

Virtual Biopsy: predicting the pathology of gliomas using multimodal deep learning

Perfusion of brain tumors In my project I assess the glioma microvasculature using intra-operative high frame rate micro-Doppler ultrasound in comparison to dynamic susceptibility contrast (DSC) MRI. Additionally I am evaluating the diagnostic accuracy of DSC-MRI using the multicenter PERISCOPE project data for distinguishing tumor progression from treatment-related changes.

While deep learning methods in radiology have allowed for improvements in the diagnosis and treatment approach of diffuse gliomas, this project explores the development of an advanced histopathology-based segmentation and grading model. By combining this with the radiology-based work from Juancito van Leeuwen, we could allow histopathomics to “help” the radiology models and explore how such a multimodal implementation can further improve detection and classification of diffuse gliomas. This will then enable us to generate histopathological “landscapes”, allowing for the transition from surgical to virtual biopsies.

Ikenna Asuzu, MSc Advisors Marion Smits, Marjolein Geurts & Brenda Leeneman

Ir. Juancito van Leeuwen, MSc

Virtual Biopsy: in silico clinical trials As the Virtual Biopsy model advances, clinical adoption becomes a realistic future goal. My project focuses on the evaluation of this model using in silico clinical trials that simulate patient pathways and clinical descision making, comparing its accuracy, treatment impact, and safety with invasive biopsy to generate robust evidence supporting safe integration and clinical acceptance in routine glioma care pathways.

Advisors Marion Smits, Stefan Klein, Pim French & Eelke Bos Virtual Biopsy: In vivo MRI glioma characterization This project aims to circumvent the need for invasive brain tumor biopsies by predicting tumor genotype and grade from MRI scans using AI. My main focus is on improving the performance of pre-existing AI models by including advanced, physiological MRI modalities during training as well as the prediction of additional genetic information.

157

Esther Warnert, PhD assistant professor 158

BENCH-TO-BEDSIDE MR IMAGING BIOMARKERS

Top Publications 2025

Esther Warnert is a medical engineer and Assistant Professor at the Department of Radiology & Nuclear Medicine. As a full-time researcher she is Principal Investigator of her research line “Benchto-bedside MR Imaging Biomarkers” in which she focusses on development and translation of novel MRI techniques to assess the brain’s physiology for clinical application. Esther obtained her MSc degree in medical engineering from the Technical University of Eindhoven in 2011. In 2015 she obtained her PhD degree from Cardiff University (Cardiff, United Kingdom) for her research into the development of non-invasive cerebrovascular MRI measurements. Esther is chair of “Glioma MR Imaging 2.0” an official Working Group of the European Society for Magnetic Resonance in Medicine and Biology, Junior Fellow of the International Society of Magnetic Resonance in Medicine, and chair of VENA, the network for women in academia at the Erasmus MC.

Wallimann P, PL Tang, F Arzanforoosh, S TanadiniLang, N Andratschke, M Smits, EA Warnert. Modeling the effect of in-plane magnetic field gradients on asymmetric spin-echo images with echo-planar imaging readout. Magnetic Resonance in Medicine 2025; 94, 2086-2099. Lefort-Besnard J, A Pron, P Clement, GJ Barker, Y Erez, MA Fernández-Seara, SB Feyzioglu, S Hartman, R Iglesias-Rey, V Keil, T Mitrović, RE Nechifor, E Ozturk-Isik, C Passarinho, I Petrušić, A Šerifović Trbalić, N Sollmann, SC Thust, G Turhan, EA Warnert, S Heunis, C Maumet. Which infrastructure can I use to share human neuroimaging data? Recommendations for EU researchers based on a survey and literature review. Aperture Neuro 2025; 2025; 512. Green KH, S van de Groep, R van der Cruijsen, EA Warnert, EA Crone. Neural correlates of well-being in young adults. Emotion 2025; 7:1677-1689.

e.warnert@erasmusmc.nl

Context

n my research line I develop, validate and apply advanced MRI techniques for innovating clinical diagnostics. I have a track record on measuring oxygen metabolism and protein content in the brain, with a strong focus on non-invasive assessment of brain tumour pathology. However, recently I am focusing on translating the advanced, physiological MRI techniques within my research line to other body parts, including the liver.

159

Research Projects: Objectives & Achievements Development & validation: Oxygen delivery to the brain Cerebral hypoxia is a devastating pathophysiological state that can occur in a plethora of diseases, including stroke, chronic hypertension, and brain tumours. The consequences of impaired delivery of oxygen to the brain are severe, from increased resistance to therapy in brain tumours such as gliomas to irreparable tissue damage and cell death in stroke. However, despite these dramatic consequences, there currently is no rapid and noninvasive assessment of hypoxia across the whole brain available in the clinic because of the complex interplay of processes involved in oxygen delivery. Physiological models assessing oxygen delivery to tissue are becoming more and more advanced, including measurements of cerebral blood flow (CBF), oxygen extraction fraction (OEF) and macro- and microvascular structure to encompass the process of oxygen delivery. Additionally, I collaborate with Prof. Dr. Marion Smits, Prof. Dr. Matthias (Thijs) van Osch (LUMC), Prof. Dr. Juan Antonio Herandez Tamames, and Dr. Dirk Poot in the supervision of Krishnapriya Venugopal, PhD candidate, on the development of novel MRI approaches to do cerebrovascular signature mapping. In the ITEM study (finished in 2024) my research team validated that vessel size imaging within brain tumours accurately reflects the microvasculute. However, validation of OEF illustrated that the use of quantitative blood oxygen-level dependent (qBOLD) to link the OEF from MRI signal deviations is not always accurate in brain tumours. We are currently exploring potential reasons for this. Together with visiting PhD candidate Philipp Wallimann, I showed that part of this inaccuracy may be explained by not appropriately correcting for magnetic field inhomogeneities. Another important aspect in investigating OEF in brain tumours is the non-invasive assessment of the CBF. In collaboration with Prof. Marion Smits our PhD candidate Anna Lavrova is therefore further exploring whether using multi-delay arterial spin labeling provides a more clear picture of brain tumour’s microvasculature that can be used to differentiate subtypes of tumours.

Development & validation: Protein measurements in the brain Malignant transformation occurs in practically all low grade diffuse gliomas. The processes leading up to tumour progression, and ultimately malignant transformation, are invisible with conventional magnetic resonance imaging (MRI) techniques used in routine clinical practice, which are mostly focused on obtaining structural T1- and T2weighted images rather than on accurately mapping in

vivo glioma biology. Hence, malignant transformation is usually established at a late stage. As aggressive cell proliferation and migration are underlying tumour growth and vessel formation, biomarkers of these processes can be used for early detection of tumour progression, eventually leading to malignant transformation. Chemical Exchange Saturation Transfer (CEST) imaging is a novel MRI technique with great potential for measuring molecular biomarkers of cell proliferation and migration within gliomas. In my longstanding collaboration with Dr. Tobias Wood (King’s College London, UK) and Dr. Thomas Booth (King’s College Hospital) I am translating amide proton transfer weighted (APTw) CEST MRI into a clinically applicable biomarker. This includes ongoing development and validation work in which we are investigating the origin of APTw-CEST MRI, by matching APTw-CEST signal from image-guided biopsies to proteomics analysis (in collaboration with Dr. Theo Luider, Department of Neurology, Erasmus MC). Additionally, I am leading a national multi-centre effort to further the use of CEST MRI as an early biomarker for detecting true tumour progression in patients who have undergone treatment for glioblastoma. This is a collaboration between the Erasmus MC and Amsterdam UMC (PI: Elsmarieke van der Giessen, PhD, MD), UMC Utrecht (PI: Evita Wiegers, PhD, MSc) and Leiden UMC (Prof. Matthias van Osch and Chloé Najac, PhD, MSc). This trial received funding from the Dutch Cancer Association (KWF) in December 2022. This trial is carried out by Laura Kemper, PhD candidate, and recruitment of patients is ongoing.

Application: advanced MRI for personalized radiotherapy of glioblastoma Together with Patrick Tang, PhD candidate, and in collaboration with Dr. Alejandra Mendez-Romero and Prof. Dr. Remi Nout of the Department of Radiotherapy) I work on integrating advanced MRI for personalized radiotherapy treatment planning of patients with glioblastoma. Followup of the patients in the MOSAIC trial is ongoing. This trial is funded by Patrick’s Mosaic fellowship of NWO and aims to illustrate how the use of biological clinical target volume, based on additional APTw-CEST, CBF and OEF MRI, can make radiotherapy more specific to active tumour whilst sparing healthy brain tissue.

Development: Oxygenation of liver tissue Tissue oxygenation is also a key feature of healthy functioning liver tissue. To address the need for a non-invasive biomarker of liver tissue oxygenation I have further developed a non-invasive, multiparametric approach for qBOLD MRI of the liver,

160

scientific report 2025 | BENCH-TO-BEDSIDE MR IMAGING BIOMARKERS

together with Inge Bosch (Msc Technical Medicine) and in collaboration with Dr Roy Dwarkasing, Prof. Dr. Juan Hernandez Tamames and GE Healthcare. I am currently developing a workflow for validation of liver qBOLD MRI through image-guided biopsies in patients with liver tumours in collaboration with Dr. Roy Dwarkasing, Dr Maarten Thomeer, Dr Kay Pieterman, Dr. Caroline Den Hoed (Gastro-enterology), Dr. Roeland de Wilde (Surgery), Dr. Michael Doukas and Dr. Thierry van den Bosch (both Pathology).

International collaboration: Glioma MR Imaging 2.0

Additional Personnel Philipp Wallimann - PhD candidate, Medical Physics at University of Zurich / University Hospital of Zurich, International exchange. Inge Bosch - MSc student Technical Medicine (Delft University of Technology. Final year research project. Xander Bos - BSc Student Informatica (Hogeschool Rotterdam) Rafael Sena - BSc Student Biomedical Engineering (University of Lisbon)

In 2025 I continued as chair for Glioma MR Imaging 2.0. A powerful tool for advancing imaging diagnostics and bringing new MRI biomarkers towards clinical application is connecting researchers and clinicians. The European network “Glioma MR Imaging 2.0” (www.glimr.eu) is doing just that, via hosting virtual, hybrid and onsite meetings and network events. It brings together over 200 researchers, clinicians and patient organisations from 30 countries and is still open to new members. A highlight of this year was our annual meeting in Marseille, France. Moreover, we have now officially become a working group of the European Society for Magnetic Resonance in Medicine and Biology.

Expectations & Directions In 2026, I will work on further consolidating the validation of advanced MRI biomarkers in patients diagnosed with brain tumours and further development of novel imaging approaches to assess cerebral and hepatic oxygenation status. Moreover, in collaboration with Dr. Simone Dalm, Dr. Tessa Brabander, and Dr. Nick Bije (Medical Oncology) I will work on a translational research line in which the MRI techniques I work on are used to explore why not all metastases of prostate cancers respond to radionuclide treatment.

Invited Lectures Esther Warnert. ‘GLiMR: the glioma and brain tumour imaging working group of the ESMRMB’. ESMRMB, Marseille, France. Oct 2025.

Funding Warnert, Esther KWF Onderzoek & Implementatie: ‘Early detection of brain tumour progression with amide proton transfer weighted MRI’. 2023-2027 Tang, Patrick NWO Mosaic Fellowship: ‘Hitting the Mark: Introducing Artificial Intelligence and state-of-the-art MRI techniques for Precision Radiotherapy of Glioblastoma’. 2022-2026

161

PhD Students Anna Lavrova, MD, M.Ed

Patrick Tang, MSc

Advisors Marion Smits & Esther Warnert

Advisors Marion Smits, Esther Warnert, Remi Nout & Alejandra Méndez Romero

Hitting the mark: Introducing artificial intelligence and state-of-the-art MRI for precision radiotherapy of glioblastoma

Advanced MR Perfusion Techniques for Glioma Characterization and Management Perfusion of glioma is an important feature for characterisation, diagnosis, treatment planning, and follow-up. In this thesis, challenges and innovations for the clinical use of dynamic susceptibility contrast and arterial spin labelling are investigated.

Glioblastomas are notorious for extensive tumor infiltration; therefore, a 1.5-cm safety margin is employed to define the target area for radiotherapy. With advanced MRI and AI comes an opportunity to more accurately define the target area for each patient and thus minimize the risk of radiationinduced side-effects.

Laura Kemper, MSc Advisors Marion Smits, Elsmarieke van der Giessen & Esther Warnert

Harmonization of APTw-CEST as a biomarker for pseudoprogression in glioblastoma A multi-center clinical trial for the harmonization of amide proton transfer (APT) weighted CEST MRI for the detection of (pseudo)progression of glioblastoma will be conducted in the Erasmus MC, UMC Utrecht, LUMC and Amsterdam UMC. The findings will be used to develop APTw-CEST as a reliable and repeatable biomarker.

162

scientific report 2025 | BENCH-TO-BEDSIDE MR IMAGING BIOMARKERS

163

Ricardo Budde, MD, PhD full professor

Alexander Hirsch, MD, PhD associate professor 164

CARDIAC IMAGING JOINT APPOINTMENT IN CARDIOLOGY

Top Publications 2025

Alexander Hirsch studied Medicine at the University of Amsterdam. He received his PhD degree (cum laude) in 2010 from the same University with a dissertation about “Clinical and functional outcomes after revascularization strategies in acute coronary syndromes”. He trained as a cardiologist at the Academic Medical Center in Amsterdam and registered as a cardiologist in 2013. He specialized further in non-invasive imaging focusing on cardiovascular magnetic resonance imaging (CMR) and obtained his level 3 certification in CMR from the European Society of Cardiology in 2014. He joined the Department of Cardiology and Radiology of the Erasmus MC in 2016. He is actively involved in scientific research and has (co)-authored over 215 publications published in peer-reviewed journals. His main research interests include the clinical application of cardiac CT and CMR in ischemic and non-ischemic cardiomyopathy. He currently supervises multiple PhD students at the Department of Cardiology and Radiology. He has been a member of the EACVI CMR certification committee since 2018 and is associate editor of the Journal of Cardiovascular Magnetic Resonance.

Budde RPJ, ME Faure, S Abbara, H Alkadhi, PC Cremer, et al. Cardiac computed tomography for prosthetic heart valve assessment. An expert consensus document of the Society of Cardiovascular Computed Tomography (SCCT), the American College of Cardiology (ACC), the European Society of Cardiovascular Radiology (ESCR), the North American Society of Cardiovascular Radiology (NASCI), the Radiological Society of North America (RSNA), the Society for Cardiovascular Angiography & Interventions (SCAI) and Society of Thoracic Surgeons (STS). J Am Coll Cardiol. 2025; 86: 1203-1230. Sharma SP, J van der Bie, AC Ziedses des Plantes, A Coenen, ML Dijkshoorn, D Bos, J Daemen, NM van Mieghem, A Hirsch, RP Budde. Diagnostic Accuracy of Ultra-High-Resolution Photon-Counting Detector Computed Tomography for Detecting Coronary In-Stent Restenosis. JACC Cardiovasc Imaging. 2025; 18: 1405-1407. van Genuchten WJ, JJ Steenhorst, GM van Tussenbroek, N van der Velde, LS Kamphuis, IK Reiss, D Merkus, WA Helbing, A Hirsch. Differences in Pulmonary Artery Stiffness Measured by CMR in Preterm-Born Young Adults With and Without Bronchopulmonary Dysplasia. Circ Cardiovasc Imaging 2025; 18: e017791.

a.hirsch@erasmusmc.nl

JOINT APPOINTMENT IN CARDIOLOGY Ricardo Budde obtained both his MD and an MSc in Medical Biology from Utrecht University. His PhD thesis concerned epicardial ultrasound in (minimally invasive) coronary artery bypass surgery. Training as a Radiologist at the University Medical Center Utrecht and registration as a radiologist was completed in 2013. He completed a fellowship in cardiovascular radiology as well as successfully passed the European Diploma in Cardiac Imaging examination in 2014. Subsequently, he joined the Erasmus MC as a staff radiologist specializing in cardiovascular and thoracic radiology. He leads the cardiovascular CT and photon counting research at EMC. He has (co)-authored over 315 publications and supervises multiple PhD students. His main research interests include photon counting CT, imaging of (prosthetic) heart valves, ischemic and structural heart disease. Ricardo is an executive board member of the European Society of Cardiovascular Radiology and Fellow of the Society of Cardiovascular Computed Tomography. He also serves as associate editor on the board of Radiology: Cardiothoracic Imaging. He is course director of the Hands-on cardiac CT course, Hands-on Photon Counting CT course and organizer of the Erasmus Photon Counting CT meeting (EPCCT). r.budde@erasmusmc.nl

Solana AB, SC Minderhoud, PA Wielopolski, JA HernandezTamames, RP Budde, WA Helbing, MA Janich, A Hirsch. Automatic failure mode evaluation using non-linear phase contrast correction to improve flow measurement accuracy in cardiovascular magnetic resonance phase contrast imaging. J Cardiovasc Magn Reson. 2025; 27: 101895.

Context

T and MRI are instrumental techniques for cardiac assessment. Technical developments have occurred at a tremendous pace over the last two decades, increasing diagnostic confidence, potential applications, and (for CT) a significant decrease in radiation dose. Cardiac CT has its role in the assessment of coronary disease including functional information (e.g., CT derived fractional flow reserve (FFR) and detailed coronary plaque assessment and quantification. Furthermore, its role in guiding cardiovascular interventions continues to expand at a rapid pace. Alongside cardiac CT, Cardiovascular Magnetic Resonance Imaging (CMR) has become an important imaging technique for patients with a wide variety of heart diseases, including ischemic heart disease, cardiomyopathies, valvular heart disease, and congenital heart disease. CMR is playing an increasingly important role in current guidelines. The cardiac imaging group represents a close collaborative effort by the departments of Radiology & Nuclear Medicine and Cardiology and consists of staff members, fellows, and PhD students from both disciplines. During 2025, we also continued our collaboration with other departments on various projects.

165

Research Projects: Objectives & Achievements Coronary Imaging and Photon Counting CT Traditionally, coronary CT angiography (CCTA) imaging has a central role in our cardiac imaging research. In 2024, we installed the second photon-counting CT (PCCT) scanner in our hospital. The improvements in spatial resolution, as well as the abilities of spectral imaging, will improve cardiac CT imaging. Important improvements in coronary imaging are seen regarding reducing calcification blooming artifacts, improved ability for quantitative assessment of coronary plaques, and better coronary luminal assessment.

We also assessed both PCCT and CT-derived FFR analyses in patients after heart transplantation and especially the temporal changes in coronary disease. Transplant patients develop accelerated coronary wall thickening and atherosclerosis (so-called cardiac allograft vasculopathy and are screened at regular intervals. CT-derived FFR can serve as an additional tool to perform a comprehensive coronary assessment in these patients. Projects: - Photon counting CT assessment (Judith van der Bie) - Photon counting CT for coronary assessment (Simran Sharma) - Imaging in heart transplant patients (Britt van Dijk)

Figure 1. Ultra-high resolution image of a stent in a coronary artery scanned on the photon counting CT scanner.

PCCT provides improved image quality of coronary CT scans using ultra-high resolution acquisition mode with slice thickness of 0.2 mm. We thoroughly evaluated the advantages and effects of this acquisition on image quality in a prospective study comparing PCCT to quantitative invasive coronary angiography for coronary stent patency assessment. The results were published in 2025. We continue our efforts in assessing CT-derived FFR as a tool to add functional information to the anatomical assessment of stenosis severity in patients with stable chest pain and patients after heart transplantation. In the multicenter randomized controlled FUSION trial, we assess CT-derived FFR in stable chest pain patients that have a >50% - <90% stenosis on CT. The primary endpoint will be the rate of unnecessary invasive coronary angiographies. Multiple Dutch hospitals participate in the FUSION trial and inclusion is now completed. Results will be presented in 2026.

Valve Disease and Endocarditis Endocarditis is a devastating disease. Prosthetic valve endocarditis is the most severe complication of valve replacement surgery and has a high mortality rate. Its diagnosis remains difficult as echocardiography is hampered by prosthetic heart valve-induced artifacts and blood cultures often being negative despite active infection. Previous research has shown that CT provides valuable information in this setting by demonstrating otherwise undetected aortic root mycotic aneurysms and vegetations that lead to changes in patient management. Positron emission tomography (PET), combined with CT, couples the detailed anatomical information of CT with the metabolic information of PET. This powerful combination is a key component of diagnosis endocarditis in the 2023 ESC Guidelines on Endocarditis. We are evaluating the additional value of ECG and respiratory gating to improve PET image quality. We published our findings on the

166

scientific report 2025 | CARDIAC IMAGING

dynamics of FDG uptake around prosthetic heart valves.

Bronchopulmonary dysplasia

More than 1000 individual patients have been discussed in our multidisciplinary “Endocarditis team”. Data on the patient characteristics, use of diagnostic techniques and diagnosis is acquired and used for scientific research.

We finished our exercise CMR study using a push-pull MRcompatible ergometer in patients with bronchopulmonary dysplasia (BPD). In total, 60 participants were included: 20 premature-born young adults with BPD and 20 prematureborn young adults without BPD. The first results were published in 2024, showing that exercise performance and cardiac response to exercise is impaired in BPD. Results from pulse wave velocity and area change measurements in the pulmonary artery were published in 2025 showing increased pulmonary artery stiffness in BPD patients.

Project: – PET CT in, and follow-up of, endocarditis patients (Eefje Dalebout)

Aortic and Valve Disease We have previously studied the role of imaging In bicuspid aortic valve study using echocardiography, CT, and CMR In the BAS study. We have extended this research with another study of congenital aortic stenosis patients (the CAS study). This study is a clinical observational study investigating the effects of congenital aortic stenosis on left ventricular function and the prevalence, pattern, and expanse of left ventricular hypertrophy, myocardial stiffness, and myocardial fibrosis. These patients underwent echocardiography with strain measurements and high frame rate echo to assess shear wave velocities and also comprehensive CMR including parametric mapping. The first results were presented in 2025 and further data will follow in 2026. Project: – Aortic and Valve Disease (Zoe Keuning)

Congenital Heart Disease Erasmus MC is an expertise center for the treatment of patients with congenital heart disease. Imaging plays an ever-increasing role in the diagnosis and follow-up of these patients. Using a dynamic phantom, we are assessing the most optimal PCCT acquisition protocols. The goal is to get the most diagnostic information at the lowest radiation dose. Next to bicuspid valve pathology, we investigate the role of CT and CMR in planning and follow-up of percutaneously implanted pulmonary valves. In the Cover study, we included patients that underwent a percutaneous pulmonary valve implantation. These patients underwent a CMR, cardiac CT, and echocardiography on the same day. The final results were published in 2025. In the Quality of Life study, the long-term cardiological and psychosocial outcomes in adults who had been operated on for congenital heart disease in early childhood are studied. Follow-up of this cohort is now more than 40 years and includes patients with a diverse spectrum of congenital heart disease from atrial septal defect to tetralogy of Fallot and transposition of the great arteries. Several results were published in 2025 and more data will follow In 2026. Project: – CT and CMR in valve disease (Marguerite Faure)

Project: – Bronchopulmonary dysplasia (Jarno Steenhorst)

Cardiomyopathy Several projects were continued and initiated in the field of non-ischemic cardiomyopathy, including the value of CMR and or CT in hypertrophic cardiomyopathy, cardiac sarcoidosis, and cardio-oncology. Furthermore, with regard to sarcoidosis, we showed that ECG and transthoracic echocardiography were of limited diagnostic value for screening for cardiac sarcoidosis but seemed to have important prognostic value as patients with normal ECG/echocardiography results who did meet the diagnostic cardiac sarcoidosis criteria had a very good prognosis. CMR/PET provided a good diagnostic yield and identified other cardiac diseases. Also, several other studies with regard to hypertrophic cardiomyopathy were continued in 2025. For example, a study about the role of cine Imaging with CMR to predict left ventricular outflow tract obstruction started, and we are also exploring the role of epicardial fat in hypertrophic cardiomyopathy patients. The results are expected to be published in 2026. Project: – Cardiomyopathy (Raveena Mangal)

Expectations & Directions Coronary CT has shifted from anatomical to functional analysis. PCCT and CT FFR will be further explored, and their role elucidated. The FUSION study is the first study to randomize patients with 50-90% stenosis to FFRct or routine care. It will provide important evidence in a randomized control trial set-up on the effect of using non-invasive FFRct to reduce the number of unnecessary invasive angiographies. The introduction of PCCT in the clinical arena has led to important improvements in cardiac CT imaging. The technique is now firmly established in our daily clinical routine. However, PCCT development is still in its infancy, and many aspects are expected to be improved even further. Ultimately, large scale patient studies will define the role of PCCT and the direction this research will take.

167

But for now, we already see improvements in cardiac imaging on a daily basis. The continuous annual followup of heart transplant patients with CCTA will provide the temporal results needed to identify the factors that are associated with (accelerated) coronary vasculopathy. The role of PCCT in assessing coronary stent patency is expected to increase as well. Data from the endocarditis database will shed more light on how to best implement PET/CT and CTA in patients suspected of endocarditis as well as identify important confounding factors. Also, longer term follow-up provides the information needed to evaluate how the diagnostic strategy is translated into quality of care. Furthermore, several projects concerning imaging data from our hypertrophic cardiomyopathy cohort and congenital heart disease studies are expected.

Ricardo Budde. ‘Clinical impact of photon counting CT in cardiac imaging’. RSNA, Chicago, USA. Dec 2025. Ricardo Budde. ‘Photon counting CT - what can it offer’? Pan London Cardiac CT User Meeting, London, UK. Oct 2025. Ricardo Budde. ‘Cardiovascular CT in children: technique and indications’. European Course of Pediatric Radiology Rotterdam, the Netherlands. Oct 2025. Ricardo Budde. ‘Imaging post-cardiac transplant’. ESCR, Glasgow, UK. Oct 2025. Ricardo Budde. ‘PCCT case readings’. NASCI, Minneapolis, MN, USA. Sept 2025. Ricardo Budde. ‘Quantitative parametric mapping: interpretation and reporting in NICM’. NASCI, Minneapolis, MN, USA. Sept 2025.

Hands-on Cardiac CT Course For many years already, we have organized the Handson Cardiac CT course in Erasmus MC. Ricardo Budde and Alexander Hirsch serve as course directors and are supported by an enthusiastic and experienced faculty including Marcel Dijkshoorn, CT technician at Erasmus MC. During 5 consecutive days, the participants read over 150 CT scans on dedicated workstations fully equipped with the latest post-processing software. This year was the 19th edition, and the course was again a huge success and completely sold out. More info can be found on our dedicated course website: www.cardiovascularimaging.nl.

Funding Budde, Ricardo, Alexander Hirsch, and consortium partners Veelbelovende zorg ZonMW: ‘Addition of FFRct in the diagnostic pathway of patients with stable chest pain to reduce unnecessary invasive coronary angiography (FUSION Study) to evaluate the role of FFRct in stable chest pain patients’. 2020-2027 Budde, Ricardo Vrienden van het Sophia: ‘Development of Photon Counting CT protocols for pediatric cardiovascular imaging’. 2023-2025 Budde, Ricardo Siemens Healthineers: ‘Evaluation of PCCT’. 2024-2028 Budde, Ricardo Bracco: ‘Assessment of coronary stents with PCCT’. 2024-2025

Invited Lectures

Ricardo Budde. ‘Clinical value and adoption of PCD-CT’. RSNA, Chicago, USA. Dec 2025.

Ricardo Budde. ‘Photon counting CT’. Lumus Imaging Conference, Melbourne, Australia. Sept 2025. Ricardo Budde. ‘Photon counting CT: How it improves clinical imaging’. I-MED Radiology Doctors’ Conference 2025, Sydney, Australia. Sept 2025. Ricardo Budde. ‘PCCT in cardiovascular imaging’. ISCT 2025, Brussels, Belgium. Sept 2025. Ricardo Budde. ‘Photon-counting cardiac CT challenge: Test your knowledge and be amazed’. SCCT, Montreal, Canada. July 2025. Ricardo Budde. ‘Photon counting CT for cardiovascular imaging’. Multidisciplinary Cardiac Imaging Conference University of Michigan, Ann Arbor, USA. July 2025. Ricardo Budde. ‘Valvular heart disease and structural interventions’. ASCI 2025, Busan, South Korea. June 2025. Ricardo Budde. ‘Practical insights into photon counting CT for cardiovascular imaging’. ASCI 2025, Busan, South Korea. June 2025. Ricardo Budde. ‘How does photon counting CT change the role of CT’? Joint Summit of the European Association of Cardiovascular Imaging (EACVI) and the European Society of Cardiovascular Radiology (ESCR), Amsterdam, the Netherlands. June 2025. Ricardo Budde. ‘Transcatheter aortic valve implantation’. ECR, Vienna, Austria. March 2025. Ricardo Budde. ‘Basic principles of cardiovascular CT’. Dutch-Caribbean Heart Days, Willemstad, Curacao. Jan 2025. Ricardo Budde. ‘FFRct’. Dutch-Caribbean Heart Days, Willemstad, Curacao. Jan 2025.

168

scientific report 2025 | CARDIAC IMAGING

Ricardo Budde. ‘Aortic dissection’. Dutch-Caribbean Heart Days, Willemstad, Curacao. Jan 2025. Ricardo Budde. ‘PCCT: the evidence this far’. Erasmus Photon Counting CT meeting, Rotterdam, the Netherlands. Jan 2025. Ricardo Budde. ‘PCCT data and PACS integration’. Erasmus Photon Counting CT meeting, Rotterdam, the Netherlands. Jan 2025. Alexander Hirsch. ‘Assessment of coronary stent patency by photon counting CT’. Erasmus Photon Counting CT meeting, Rotterdam, the Netherlands. Jan 2025.

Nov 2025. Alexander Hirsch. ‘Cases with the experts: cardiac magnetic resonance and structural heart disease (3x)’. EACVI, Vienna, Austria. Dec 2025. Alexander Hirsch. ‘Advanced cardiac magnetic resonance: mapping techniques’. EACVI, Vienna, Austria. Dec 2025.

Additional Personnel Mohamed Attrach – MD, Cardiovascular Radiologist

Alexander Hirsch. ‘CT in endocarditis: A practical handson workshop’. SCCT Global, Amsterdam, the Netherlands. Feb 2025. Alexander Hirsch. ‘NAEOTOM Alpha Class Face-off: A battle of the best Photon-counting cases’! SCCT Global, Amsterdam, the Netherlands. Feb 2025. Alexander Hirsch. ‘Lesion-specific physiology + quantified plaque to inform decision-making’. SCCT Global, Amsterdam, the Netherlands. Feb 2025. Alexander Hirsch. ‘The role of echocardiography and MRI in cardiac amyloidosis’. 5th Multidisciplinary Symposium on Systemic Amyloidosis, Utrecht, the Netherlands. March 2025. Alexander Hirsch. ‘Cardiac MRI for every cardiomyopathy’. CCR, Rotterdam, the Netherlands. June 2025. Alexander Hirsch. ‘Why Coronary Anatomy Matters: The Role of Coronary CT in Coronary Syndromes’. Cardiology conference of the Journee Wallonne de L’Imagerie Cardiaque, Liege, Belgium. March 2025. Alexander Hirsch. ‘Diagnostic pathway for stable chest pain’. Cardiac Imaging Masterclass. London, United Kingdom. May 2025. Alexander Hirsch. ‘Latest progress with photon-counting CT’. Workshop on AI-Powered Strategies towards Advancing PCI, Leiden, the Netherlands. May 2025. Alexander Hirsch. ‘Role of CT-FFR within CLEAR-CAD’? CLEAR-CAD investigators meeting, Venlo, the Netherlands. Sept 2025. Alexander Hirsch. ‘How to do: T1/T2/T2* mapping’. Amsterdam Cardiac MR Course, Amsterdam, the Netherlands. Oct 2025. Alexander Hirsch. ‘Infiltrative cardiomyopathies’. Amsterdam Cardiac MR Course, Amsterdam, the Netherlands. Oct 2025. Alexander Hirsch. ‘CMR technique, protocol and indications’. Pediatric Cardiovascular Imaging, European Course Pediatric Radiology, Rotterdam, the Netherlands.

169

Post-doc

Annick Weustink MD, PhD Cardiovascular Nuclear Imaging My research focuses on cardiovascular nuclear imaging, with a particular interest in myocardial perfusion imaging, cardiovascular inflammation, and cardiomyopathies. The aim is to improve the diagnosis and management of cardiovascular disease through the use of quantitative imaging biomarkers. Current work centres on quantitative myocardial perfusion imaging using the PET tracer ¹⁸F-flurpiridaz in patients with coronary artery disease and heart transplant recipients. These studies investigate how PET-derived measures of myocardial blood flow and myocardial flow reserve can support patient management and assessment of treatment response. Together, they contribute to the growing role of quantitative PET imaging in cardiovascular care. In parallel, ongoing projects explore nuclear imaging of cardiovascular inflammation and cardiomyopathies, including infective endocarditis, cardiac sarcoidosis, and myocardial fibrosis. This work examines how advanced imaging can improve disease characterisation and support diagnosis and treatment monitoring. Future research will focus on the application of long axial field-of-view (LAFOV) PET/CT in cardiovascular disease, particularly in patients with advanced heart failure. By enabling comprehensive assessment of perfusion and organ function, this technology may offer new opportunities for disease assessment and risk stratification. Together, these studies aim to further define the role of cardiovascular nuclear imaging in patient care.

170

scientific report 2025 | CARDIAC IMAGING

PhD Students Britt van Dijk, BSc

Eefje Dalebout, MSc

Advisors Ricardo Budde, Olivier Manintveld, Daniel Bos & Rudolf de Boer

Advisors Jolien Roos - Hesselink, Ricardo Budde, Jolanda Kluin & Alexander Hirsch Cardiac imaging in infective endocarditis Imaging is one of the cornerstones in diagnosis and management of infective endocarditis. This PhD trajectory will focus on the role of imaging techniques such as PET-CT and advanced CT techniques in patients with (suspected) endocarditis and its additional value in diagnosis and clinical management.

Cardiac Allograft Vasculopathy after Heart Transplantation MRI-detected subclinical joint inflammation is a promising predictor of progression from clinically suspect arthralgia to rheumatoid arthritis. Our project aims to optimize the Dixon MRI sequence and enable automatic detection of subclinical Inflammation in the hands. This will Improve the feasibility of MRI for routine rheumatology practice.

Simran Sharma, MD, MSc Advisors Ricardo Budde, Alexander Hirsch & Nicolas van Mieghem

Marguerite Faure, MD, MSc Advisors Ricardo Budde & Alexander Hirsch

FFRct and Coronary Artery Disease Fractional Flow Reserve derived from coronary computed tomography (FFRct) analysis is a noninvasive technique that uses the CCTA images as a basis for complex software-based calculations and modelling to provide additional functional information based on the anatomical CCTA images. To investigate the impact of adding the FFRct analysis to the diagnostic pathway of stable chest pain patients, we have set up the FUSION study, a national, multicentre, randomised controlled trial.

CT and MRI Imaging of prosthetic heart valves This project focusses on imaging bioprosthetic valves with CT and MRI. In particular patients that underwent a percutaneous pulmonary valve implantation (PPVI). CT seems a reliable tool for risk assessment of coronary artery compression. There seems no relevant change in RVOT to coronary distance and coronary lumen diameter after PPVI. Conduit expansion does not seem to affect the relationship between the pulmonary trunk and coronary arteries after implantation. In a second part of the project we are now evaluating flow patterns after PPVI with 2D and 4D flow MRI.

171

Ivo G Schoots, MD, PhD associate professor 172

ABDOMINAL IMAGING

Top Publications 2025

Ivo Schoots completed his professional radiologic training in 2012 at the Academic Medical Center, Amsterdam after a certified clinical fellowship in abdominal radiology. In 2004 he received his PhD on the ‘crosstalk of coagulation and inflammation in ischemia and reperfusion mechanism’ at the University of Amsterdam (Surgery & Internal Medicine). He performed a postdoctoral research fellowship at Harvard Medical School, Department of Molecular & Vascular Medicine, BIDMC, Boston, MA/USA. His research is now focused on oncologic abdominal imaging, in particular on developments in prostate cancer MR imaging, imaging biomarkers, and clinical decision modeling. He participates in international and national working groups on prostate cancer imaging and is involved in the clinical implementation of prostate MRI in combination with risk stratification.

Prinsen Anne MA, IM Wagensveld, DF Osses, PM Groenendijk, FM Zijta, MB Busstra, E Rociu, JO Barentsz, JP Sedelaar, B Arbeel, T Roeleveld, R Geenen, I Koeter, S van der Meer, VC Cappendijk, S Klaver, H van der Lely, T Wolters, MR Leter, HG van der Poel, S Heijmink, J Immerzeel, JA Leijte, JG van Roermund, RL Miclea, E Planken, AN Vis, IJ de Jong, J Tijsterman, D Wolterbeek, A Claessen, J Nederend, CH Bangma, MJ Roobol, S Remmers, RC van den Bergh, DM Somford, IG Schoots. Extending Current Guideline-recommended Prostate-specific Antigen Density Thresholds for Improving Benefit/ Harm Balance in Risk-based Magnetic Resonance Imaging-directed Prostate Cancer Diagnosis, from the Community-based Multicentre MR-PROPER Study. Eur Urol Oncol. 2026; 9:571-580.

i.schoots@erasmusmc.nl

Veenstra MM, S Veldhuijzen van Zanten, E Vegt, IA Boere, HJ van Beekhuizen, J Nonnekens, FA Verburg, MG Thomeer. Emerging Radiopharmaceuticals Beyond FDG for Ovarian Cancer: A Review of Advances in Nuclear Medicine. Cancer Med. 2025; 14:e71167. Schoots IG, AR Padhani AR. A Paradigm Shift Toward Population-Based MRI-Targeted Prostate Cancer Screening. JAMA Oncology 2025; 11:828-30.

Context

ur Abdominal Imaging research is focused on several organs, in the lower abdomen as well as in the upper abdomen. Some of our researchers have developed particular skills and interest in advanced MR imaging of the prostate or cervix within the clinical decision-making algorithms, others are more dedicated to liver and colorectal cancer imaging. We are very happy that in the collaboration with our clinical partners in Surgery, Gastroenterology, Urology, Gynecology, Medical Oncology, and Radiotherapy is structurally bearing fruit by attracting funding and generating some promising scientific output. We are convinced that in the coming years some of these areas of research and probably even other topics in abdominal imaging will gain more interest and become increasingly successful.

173

Research Projects: Objectives & Achievements Liver cancer imaging As part of the largest liver center in the Netherlands, continuing efforts are being made to improve the capabilities of non-invasive diagnostics of liver tumors. Our study projects on hepatocellular cancer (HCC) have focused on the use of different diagnostic modalities (ultrasound (US), computed tomography (CT), and MRI) for the detection and subsequent decision-making of HCC. This tumor is the tenth most common malignancy and sixth cause of cancer-related deaths in men. Our referral center for hepatobiliary diseases is a screening center for HCC in hepatitis and cirrhosis. Currently, several projects are focused on further optimizing the screening and diagnostic pathways, with shorter and more (cost) efficient MRI screening protocols. Together with the biomedical imaging group of our department, we focus on multicentric radiomics platform for liver lesion phenotyping using MRI data from different liver centers in the Netherlands (Dutch Hepatic Cancer Group) and around the world (Liver Artificial Intelligence (LAI) consortium). Efforts are made to find generalizable applications for liver tumor classification and prognosis. Furthermore, together with our interventional radiologists we investigate several minimal invasive techniques in treating liver cancers. Moreover, international collaborations continue to expand, including joint AI projects on pancreatic cancer imaging. These collaborations focus on validation of AI-based imaging tools in European patient cohorts and contribute to strengthening the international translational research network of the department.

Figure 1. Liver cirrhosis due to chronic hepatitis C infection, demonstrating an early HCC (<2 cm (LI-RADS 5, white arrows). Images (a) and (b) represent T2-weighted (T2W) and diffusionweighted imaging (DWI), respectively, and are part of the Short MRI Surveillance (SMS) protocol for HCC surveillance. Images (e) and (f) are contrast-enhanced MRI acquisitions.

Figure 2. Hepatocellular adenoma (HCA) demonstrating atypical imaging features, including lesion inhomogeneity and central necrosis. Due to its large size and atypical appearance, the lesion was surgically resected. Histopathological analysis revealed a β-catenin–mutated HCA, which is considered a high-risk subtype because of its increased potential for malignant transformation. Ongoing research within our group focuses on the imaging characteristics and clinical outcomes of known liver lesions with atypical imaging findings.

Prostate cancer imaging Prostate cancer is the most common malignancy and leading cause of cancer-related deaths in men. Optimization of diagnosis and treatment of prostate cancer is of great social importance. Prostate cancer, however, is a heterogeneous disease with aggressive and non-aggressive tumors. Over the past decade prostate MRI has impacted diagnostic management substantially by identifying mainly aggressive tumors that need treatment. The advanced diagnostic strategies improve diagnostic accuracy, needing less biopsy cores, and navigate to better therapy choice with subsequent lower morbidity, increased quality of life and at lower costs. Our projects on imaging and image-guided strategies of prostate cancer address the need to reduce over-diagnosis, over-biopsy and overtreatment of prostate cancer, due to today’s less than ideal screening tests. Our results of several trials have led to impactful guideline changes, and to the renewed question whether to screen for prostate cancer with MRI or not.

Figure 3. Cancer detection rates (CDR) for GG= 1, GG ≥ 2, benign TZ lesions with their corresponding 95% confidence intervals and trend lines for PI-RADS 1-5 with additional categories “2 + 1” and“3 + 1” separated. From Agrotis, Schoots, et al. Detection rates for

174

scientific report 2025 | ABDOMINAL IMAGING

prostate cancer using PI-RADS 2.1 upgrading rules in transition zone lesions align with risk assessment categories: a systematic review and meta-analysis. Eur Radiol 2025.

Being part of several national and European initiatives on prostate cancer diagnosis and population-based screening, our projects focus on biomarker imaging, quality standards, risk assessment, and AI-assistance and stand-alone requirements, to lower the harms of overbiopsy, over-diagnosis and over-treatment.

Figure 4. Visual summary of accelerated MRI in prostate imaging using conventional and deep learning reconstruction (DLR). Furthermore, potential use cases, challenges, and future questions for the implementation of accelerated prostate imaging protocols. From Urpsrung, Schoots, et al. Prostate MRI Using Deep Learning Reconstruction in Response to Cancer Screening Demands-A Systematic Review and Meta-Analysis. J Pers Med 2025.

Rectal cancer imaging Colorectal cancer is the third most common malignancy and third cause of cancer-related deaths in men and women. Our projects mainly focus on the diagnosis and treatment management of local recurrent rectal cancer, centralized in a few centers in the Netherlands. Dedicated MRI is an integral part of the work-up and important for clinical decision making. Our research aims to develop and validate modern MR techniques for better selection and follow up of complete responders after preoperative chemoradiotherapy of (recurrent) rectal cancer. Furthermore, imaging biomarkers with tumor texture analyses may help predict surgical outcome and long-term prognosis.

Cervical cancer imaging Cervical cancer is the fourth most common cancer among females, and the fifth cause of cancer-related death in women. Our research has changed guideline

Figure 5. Consumption and supply-based hypoxia (CSH) model application workflow. Lesion is depicted in a purple contour. H&E: Hematoxylin and eosin staining, T2-w: T2 weighted image, ADC: Apparent diffusion coefficient map, fBV: fractional blood volume map, HL: Hypoxia level (pixel-level), HFDWI: Hypoxia fraction (lesion-level). From Fernandez, Schoots, et al. Hypoxia and cribriform growth in prostate cancer - establishing a link via MRI. BMC Med Imaging 2025.

175

recommendations with cervix MRI being the hallmark for primary diagnostic staging. The diffusion weighted imaging (DWI) showed to be a critical part of the MRI protocol, increasing the detection of residual tumor following chemoradiation significantly. Based on an Erasmus MC lead collaborative prospective multicenter study, international guidelines on cervical cancer have adopted the recommendations. Furthermore, clinical impact studies on other gynecological tumors are progressing (e.g., ovarian cancer).

Expectations & Directions Our research promotes technological advancement in clinical application to guide state-of-the-art diagnosis and to further drive the improvement of (imaging-based) therapeutic approaches. Our research contributes to a better understanding of the biological characteristics of cancer and non-cancerous tissue. MR imaging biomarkers are being explored and validated for personalized decision-making and for prediction of treatment efficacy. The development and validation of these methods is research in joint cooperation with industrial partners. Association of imaging traits with gene expression profiles will certainly be an area of interest, in close correlation with other Erasmus MC departments and (inter)national research groups. Developments on nuclear imaging techniques (o.a. PSMA, FAPI) will be explored to correlate morphological and functional MR images to PET CT images. We will further expand our imaging expertise and apply novel imaging methods to improve our understanding of the genetic and disease development in abdominal oncology. We will initiate new projects in the field of oncological imaging analysis, with machine learning and deep learning techniques. We will continue to increase our research efforts into investigating clinically related oncological research questions. We intend to explore strategies to improve diagnostic accuracy, reduce the harms of overdiagnosis, and improve therapeutic decisionmaking, thereby lowering morbidity, increasing quality of life, and reducing costs.

diagnostisch traject prostaatkanker middels MRI’. 2022 – 2026 Dwarkasing, Roy, Francois Willemssen, Rob de Man, Bart Takkenberg, and Carine Uyl-De Groot KWF implementation grant: ‘Validation of a Short and effective MRI Surveillance protocol for hepatocellular carcinoma screening in practice’. 2022 – 2026 Lambregts, Doenja, Ritse Mann, and Ivo Schoots KWF implementation grant: ‘Towards clinical implementation of novel diagnostic tools in oncologic imaging: an opensource web-based reviewing infrastructure for validation studies’. 2023 – 2027 Aben, Katja, Berdine Heesterman, Daniela Oprea, Pim van Leeuwen, and Ivo Schoots KWF implementation grant: ‘ADMINISTRATE (Advanced Diagnostic Modalities in ImagiNg Impacting on diagnosiS, TReatment And paTient outcomE) on prostate cancer’. 2023 – 2026 Thomeer, Maarten, et al. KWF clinical research Grant: effect of FAPI PET-CT on management in patients with potentially resectable biliary tract cancers: prospective multicenter study and cost-effectiveness analysis. 20232027

Highlights José Castillo Tovar defended his thesis Computer Aided Diagnosis for Prostate Cancer Classification on Multiparametric MRI successfully on 11th of November 2025.

Additional Personnel Demi Huijgen – PhD student Petriatic Surgery Erasmus MC Dr. Kay Pieterman MD, PhD – Post doc researcher, Interventional Radiologist, Erasmus MC Frederik Hartmann Ruben Niemantsverdriet

Funding

Angela Amirabile – Humanitas Research Hospital Italy

Pensky, Bagdi Ulas, Wallace Bolan, Gonda Hecht, Marco Bruno, and Ivo Schoots NIH grant: ‘Deep learning methods for characterization of pancreatic cysts (Cyst-X project)’. 2021 - 2025 Schoots, Ivo, and Uulke Van der Heide KWF clinical implementation grant: ‘PROCESS study: PROstate Cancer - Expansion of Surveillance Selection criteria with MR imaging’. 2021 – 2025

Laurentine Rietveld – Masterproject Yorvanny Salazar- Masterproject Jelle Ippel- Masterproject Wenny Chen- Masterproject

Van den Bergh, Roderick, Rik Somford, and Ivo Schoots SKMS project/ZonMW: ‘Evaluatie en optimalisatie

176

scientific report 2025 | ABDOMINAL IMAGING

Assistant Professor

Roy Dwarkasing, MD, PhD Email r.s.dwarkasing@erasmusmc.nl

Roy was appointed Assistant Professor in 2025. After completing his radiology training at Erasmus MC Rotterdam, he obtained his PhD degree at Erasmus University Rotterdam in 2016. He specializes in abdominal imaging and is principal investigator in imaging surveillance and diagnosis of hepatocellular carcinoma. His research focuses on imaging surveillance, optimized diagnosis, and AI-driven prediction modelling in liver and pancreatic tumors, aiming to improve diagnostic accuracy and personalized clinical decision-making. Our research group focuses on imaging surveillance, diagnosis, and outcome prediction in liver and pancreatic tumors. A major ongoing project is the nationwide multicenter KWF-funded SMS study, evaluating a non-contrast short MRI surveillance protocol for hepatocellular carcinoma (HCC) in high-risk patients compared with ultrasound surveillance. The study represents one of the largest prospective abbreviated MRI surveillance studies for HCC in Europe and continued its final phase of execution during 2025 (PhD- student Céline van de Braak, MSc). Additional research focuses on optimization of MRI protocols for accurate characterization of primary liver tumors and atypical liver lesions, including hepatocellular adenoma (HCA), focal nodular hyperplasia (FNH), atypical hemangioma, and cholangiocarcinoma (CCA). These projects aim to reduce diagnostic uncertainty and improve MRIbased diagnostic pathways (PhD student François Willemssen, MD). The group is also increasingly active in imaging biomarkers, interventional imaging, radiomics, and artificial intelligence (AI). Current projects evaluate AI-driven prediction models and imaging biomarkers for treatment response in HCC, including patients treated with selective internal radiation therapy (SIRT). In parallel, the nationwide Inter-HCC consortium is being developed to integrate imaging, clinical data and AI prediction models for clinical decision support. International collaborations continue to expand, including a Taiwan–Netherlands research initiative on AI-assisted detection of small pancreatic cancers using the PANCREASaver AI tool. Together, these projects build a coherent imaging-driven research line ranging from surveillance and diagnosis to outcome prediction and personalized clinical decision support. Future research aims to explore other PET tracers that may offer greater specificity for distinct pain pathways, further advancing diagnostic capabilities in this field.

177

Assistant Professor

Maarten Thomeer MD, PhD Email m.thomeer@erasmusmc.nl

Maarten was appointed Assistant Professor in October 2025. He studied Medicine and completed his radiology training at the Catholic University of Leuven, Belgium. He obtained his PhD in 2018 at Erasmus University Rotterdam. Since 2002, he has been a member of the abdominal radiology staff, where he is responsible for the development, optimization, and quality assurance of MRI protocols for abdominal imaging. Dr. Thomeer serves on the management board of the LAI Consortium. In addition, he is an active member of the Dutch Benign Liver Tumor Group, Dutch Hepatocellular and Cholangiocarcinoma Group and the Dutch FAPI Working Group. In 2025, our research focused on MRI and PET imaging of abdominal tumors. We published a multicenter study identifying imaging features that differentiate β catenin–mutated adenomas from FNH—lesions with similar MRI appearance but distinct clinical relevance. Findings were presented at ESGAR 2025; external validation is ongoing. The LAI consortium is operational, building a large liver MRI dataset (>3,000 cases). In parallel, multiple large language models are being evaluated for explainable liver tumor classification, with results expected in 2026. Iinclusion continues in the FAPI-Chol PET study, with promising preliminary data. Mara Veenstra contributed a review on novel PET tracers and real world studies showing FAPI PET/CT outperforms FDG in abdominal cancers.

Figure: CT and FAPI PET/CT of a patient scheduled for resection of perihilar cholangiocarcinoma. Preoperative staging with state of the art CT and MRI showed no suspicious lymph nodes or distant metastases. Following inclusion in the FAPIchol study, the patient underwent FAPI PET/CT (b), which revealed multiple highly suspicious peritoneal deposits not detected on CT or MRI. Retrospective review of the CT images (a) demonstrated only subtle, non specific abnormalities in the regions corresponding to the areas of intense FAPI uptake. These additional findings led to a change in clinical management: instead of surgical resection, the patient was treated with systemic chemotherapy.

178

scientific report 2025 | ABDOMINAL IMAGING

PhD Students Mara Veenstra, MSc

Céline van de Braak, MSc

Advisors Maarten Thomeer, Erik Verburg & Rutger-Jan Swijnenburg

Advisors Aad van der Lugt. Rob de Man, Roy Dwarkasing & Daniel Bos

FAPI PET/CT for cholangiocarcinoma

Validation of Short MRI Surveillance (SMS) protocol for hepatocellular carcinoma (HCC) screening in practice

We believe that additional preoperative [18F] F-FAPI PET/CT for patients suffering from proximal cholangiocarcinoma can detect metastases that cannot be found using CT and MRI, and that futile surgeries can be avoided using this new work-up.

An abbreviated MRI protocol for HCC surveillance is validated in a high-risk patient population and compared with biannual ultrasound surveillance, the current standard of care. To our knowledge, this is the first prospective head-to-head comparison of these surveillance strategies and may improve detection of early-stage HCC and, consequently, patient survival.

François Willemssen, MD, MSc Advisors Roy Dwarkasing & Daniel Bos Dr. Willemssen is responsible for the radiological contribution in the Multidisciplinary Liver Tumor Board, participates in the research committee of the Dutch Hepatocellular and Cholangiocellular Group (DHCG), is involved in the revision of two national guidelines, Cholangiocarcinoma and Hepatocellular carcinoma, and contributes to several research projects on liver imaging. His PhD project focuses on MRI applications in liver tumor diagnosis and on optimization of imaging strategies for characterization of atypical liver lesions.

Neslisah Seyrek, MD, MSc Advisors Ivo Schoots & Arno van Leenders

The value of mpMRI on prediction of cribriform growth in prostate cancer Assesment of the presence of cribriform growth plays a crucial role in risk stratification and active survellience eligibilty of prostate cancer patients, especially in intermediate risk group. Adverse imaging parameters derived from mpMRI may contribute in prediction of cribriform presence and enhance the presicion of risk assesment of men with intermediate risk prostate cancer.

179

Edwin HG Oei, MD, PhD full professor 180

ADVANCED MUSCULOSKELETAL IMAGING RESEARCH ERASMUS MC (ADMIRE) Top Publications 2025

Edwin Oei is a Full Professor of Musculoskeletal Imaging, and Section Chief of musculoskeletal radiology in Erasmus MC’s Department of Radiology & Nuclear Medicine. He obtained his medical degree in 2004 and his PhD on MRI for traumatic knee injury in 2009, both from Erasmus University Rotterdam. He also holds an MSc in Clinical Epidemiology from the Netherlands Institute for Health Sciences. Dr Oei is the principal investigator of musculoskeletal imaging research and engages in many academic activities including supervising 15 PhD students, lecturing, board and committee memberships, and refereeing for various journals. He is a recent Past-President of the European Society for Magnetic Resonance in Medicine and Biology (ESMRMB) (2023-2024).

Booij R, P de Klerk, E Tesselaar, M Woisetschläger, A Brandts, M Olsthoorn, J van Oldenrijk, K Bos, J Schilcher, EHG Oei. Assessment of bone-implant interface image quality for in-vivo acetabular cup implants using photon-counting detector CT: Impact of tin pre-filtration. Eur J Radiol Open. 2025; 27;14:100646. Santini F, M Pansini, X Deligianni, ME Caligiuri, EHG Oei. ESR Essentials: advanced MR safety in vulnerable patients-practice recommendations by the European Society for Magnetic Resonance in Medicine and Biology. Eur Radiol. 2025; 35:17851793.

In 2013, Edwin Oei spent a one year research sabbatical as a Visiting Assistant Professor in the Joint and Osteoarthritis Imaging with Novel Techniques (JOINT) lab of the Department of Radiology of Stanford University, CA/USA. Dr. Oei is also the Clinical Research Focus Group leader in the Department of Radiology & Nuclear Medicine, and the principal coordinator of the Academic Center for Molecular and Cellular Imaging at Erasmus MC.

van der Heijden RA, D Tamada, L Mao, J Rice, SB Reeder. Relaxivity and In Vivo Human Performance of Brand Name Versus Generic Ferumoxytol. Invest Radiol. 2025; 60:319-327.

e.oei@erasmusmc.nl

Context

maging is key in the study of musculoskeletal diseases. More sensitive and accurate imaging techniques are needed to advance our understanding of the development, mechanisms, and subtypes of musculoskeletal disorders. This is especially important for disorders that are common and have large impact on patients and the society, such as osteoarthritis, osteoporosis, sports injuries, and chronic musculoskeletal pain. New imaging biomarkers based on CT, MRI, ultrasound and nuclear imaging can also facilitate early diagnosis and development of targeted therapies. Before they can be applied routinely, new imaging techniques need to be optimized and validated. When applied in clinical research studies, novel, quantitative, imaging tools can provide additional imaging biomarkers. In large populationbased studies, imaging is essential to characterize the development and aging of the musculoskeletal system. To analyze large imaging datasets, efficient image analysis methods, incorporating radiomics and artificial intelligence, are crucial.

181

Research Projects: Objectives & Achievements Optimization and validation of advanced imaging techniques for common musculoskeletal diseases

•

PET-MRI for chronic musculoskeletal pain (Marijn Mostert and Elisa Maijoor)

•

Quantitative ultrasound to assess bone marrow lesions (Ilaria Orsi)

Application of advanced imaging techniques in clinical studies

Figure 1. Ultra-high resolution photon counting CT image of the elbow displaying an osteochondral defect of the capitellum of the humerus.

This research line addresses the optimization and validation of novel imaging techniques based on CT, MRI, ultrasound and nuclear imaging. We successfully validated quantitative MRI techniques to assess altered processes or composition in musculoskeletal tissues such as cartilage, meniscus and tendon, against tissue references and clinical outcomes. We also validated advanced, contrastenhanced ultrasound and shear-wave elastography in osteoarthritis and tendinopathy. More recently, photon counting CT has become of great interest as it has shown promise to characterize bone disorders such as fractures and osteoporosis better than current methods because of its ultra-high resolution (Figure 1). Validation for routine clinical use also includes assessment of additional value for patients and clinicians. We are currently conducting a randomized controlled diagnostic trial to study the value of PET-MRI for chronic musculoskeletal pain . In this research line, we collaborate closely with the Physics In MR and CT Technology Groups , Biomedical Imaging Group Rotterdam, Departments of Orthopedics and Sports Medicine and Pain Medicine (Frank Huygen), TU Delft (Guillaume Renaud) and external partners (University of Wisconsin, Stanford University, Mayo Clinic, GE Healthcare, and Siemens Healthineers).

A strength of our research is the multidisciplinary embedding and collaboration with many clinical research groups, enabling the application of novel imaging techniques in clinical studies, providing additional imaging biomarkers. Previously, we successfully applied advanced quantitative MRI techniques for cartilage composition (dGEMRIC, T1rho-, and T2-mapping) in several clinical studies on knee osteoarthritis. In some studies we also applied perfusion imaging with dynamic contrast-enhanced MRI (DCEMRI) (Figure 2). We successfully implemented ultrashort echo time (UTE) MRI and shear-wave elastography (SWE) ultrasound in the largest randomized controlled trial of patients with patellar tendinopathy (study funded by General Electric and National Basketball Association). We recently also introduced a short MRI scan of both hands without the need for contrast agents in a cohort of patients with early clinically suspected arthralgia. These projects are conducted in collaboration with the departments of Orthopedics and Sports Medicine (Duncan Meuffels, RobertJan de Vos, Koen Bos, Max Reijman, Denise Eygendaal, Sita Bierma-Zeinstra), Rheumatology (Annette van der Helm-van Mil, Pascal de Jong). Current projects: •

Advanced MRI and ultrasound in a randomized controlled trial of patellar tendinopathy (Jie Deng)

•

Short Dixon MRI for hand arthralgia (Sanne Boeren, Daniek van der Kaaij)

Current projects: •

Photon counting CT (Thom van der Laan, PHD student with Physics in CT group)

Figure 2: Dynamic contrast-enhanced MRI (DCE-MRI) based perfusion map of the synovium in the knee.

182

scientific report 2025 | ADVANCED MUSCULOSKELETAL IMAGING RESEARCH ERASMUS MC (ADMIRE)

Musculoskeletal imaging in population studies We are actively involved in the two large population studies in Erasmus MC – the Rotterdam Study and Generation R. In the Rotterdam Study among middle-aged and elderly. Our research focuses on studying determinants and imaging aspects of osteoarthritis and osteoporosis. In the Generation R cohort of children and adolescents we play a leading role in studying development, growth and abnormalities of the hip, knee, and spine on focused rapid MRI scans as well as body composition on whole-body MRI (Figure 3). For such large datasets, it is necessary to make use of automated image analysis tools, which we develop together with the Biomedical Imaging Group Rotterdam (Dr. Stefan Klein and Dr. Jukka Hirvasniemi). In both projects, we collaborate closely with researchers from the departments of General Practice (Sita Bierma-Zeinstra, Marienke van Middelkoop), Orthopedics and Sports Medicine (Tom Piscaer), Internal Medicine (Joyce van Meurs), and Public Health (Hein Raat).

enables precision diagnosis of joint load in the context of musculoskeletal disorders affecting biomechanics, by combining biomechanical data with advanced imaging using (PET-)MRI. A first study on knee osteoarthritis with this integrated approach was started at the end of 2025. This joint initiative with researchers from TU Delft (Ajay Seth, Jaap Harlaar) marks one of several projects, largely conducted within the Flagship “Healthy Joints” as part of the “Health & Technology by Convergence” initiative between Erasmus MC Rotterdam, TU Delft and Erasmus University Rotterdam, in which our group plays a prominent role. Current projects: • Biomechanical precision diagnostics in osteoarthritis (Niels Dur) • Healthy Loading to combat osteoarthritis (LoaD project) (Wouter Schallig)

Current projects: • Hip shape in children (Desirée de Vreede and Mirthe Kamphuis) • Spine abnormalities in adolescents (Marleen van den Heuvel) • Knee shape and abnormalities in adolescents (Rosemarijn van Paassen and Laura Kemmeren)

Figure 4. The MOBI lab setup

Miscellaneous projects

Figure 3. 3D shape analysis of the femur based on automated segmentation of MRI scans in the Generation R cohort, showing different shape modes and the variations they explain.

Biomechanics and imaging An important new direction in our research is the integration of (PET-)MR imaging with biomechanical measurements in the new Motion Biomechanics & Imaging (MOBI) lab, opened In October 2024. This is one of the first clinical laboratories in Europe that combines traditional motion capture with fluoroscopy imaging during movement (Figure 4). This work

We study various other topics in the field of musculoskeletal imaging. We are involved in many clinical studies on osteoarthritis using ”conventional” radiographic or MR imaging with semi-quantitative osteoarthritis grading. Examples include studies on specific risk factors for osteoarthritis such as overweight, anterior cruciate ligament rupture, or menopause, in collaboration with the departments of Orthopedics and General Practice. We also completed a randomized sham-controlled clinical trial on the efficacy of genicular artery embolization with follow-up using advanced (DCE-)MRI. In another project, we assess hip and groin disorders in athletes, utilizing generated CT-like images from MRI (Figure 5). Together with the Value-based Imaging group (Jacob Visser) we conduct several clinical validation studies of artificial intelligence algorithms, e.g. for fracture detection or osteoarthritis grading on radiographs, collaborating with industrial partners, and with the BIGR group we work on more fundamental AI and radiomics projects, e.g. on musculoskeletal tumors. We also fulfil a consulting or supportive role in many other studies primarily conducted by other groups at Erasmus MC and internationally (e.g. University of Queenland, LaTrobe, Lund, and Stanford Universities). Current projects: • Genicular artery embolization for knee osteoarthritis (Tijmen van Zadelhoff)

183

• • •

Van der Heijden, Rianne NWO/ZonMw Veni talentprogram: ‘Towards better care for patients with chronic low back pain using advanced imaging’. 2024-2029

Imaging of hip and groin disorders in athletes (David Hanff) Artificial intelligence for musculoskeletal radiology (Sybren van Hal) Trustworthy artificial intelligence for musculoskeletal tumors (Xinyi Wan)

Oei, Edwin NWO Research along routes by Consortia (NWA-ORC): ‘Healthy Loading to combat osteoarthritis: Leveraging molecular variations in load bearing capacity for individualized movement aDvice: The LoaD project’. Main applicant: G.J.V.M. van Osch (Orthopedics). 2022-2030 Van der Heijden, Rianne Stichting Erasmus MC Pijnfonds: ‘Richting betere zorg voor patienten met chronische lage rugpijn met geavanceerde beeldvorming’. 2024-2027 Oei, Edwin NWO ROBUST program and General Electric Healthcare: ‘Innovation Center for Artificial Intelligence (ICAI) lab Trustworthy AI for Magnetic Resonance Imaging’. 20232026 Van der Heijden, Rianne AASPT New Investigator Grant, Freedom of movement Grant, and General Electric (GE) sponsored project : ’Identifying Factors Associated with Early Signs of Osteoarthritis in Former Collegiate Athletes after Anterior Cruciate Ligament Reconstruction’. 2022-2026

Figure 5. CT-like images of the hips and pelvic skeleton generated from MRI using a zero echo time (ZTE) technique.

Expectations & Directions We have planned an expanding research activity in PETMRI, both for assessment of knee osteoarthritis (integrated with the MOBI lab) and musculoskeletal pain, which will be directed mainly by Rianne van der Heijden. There will also be exciting new developments in photon counting CT for musculoskeletal applications which will need to be validated. In image analysis there will obviously be an increased activity in artificial intelligence, both in terms of validating and determining additional value of clinical AI algorithms as in the development and application of research oriented algorithms. Finally, we expect an growing role in Generation R with the availability of more MR imaging biomarkers across different joints and with regard to body composition.

Funding Oei, Edwin General Electric Healthcare: ‘Pinpointing the source of chronic pain and therapy response with wholebody 18F FDG-PET/MRI’. 2021-2025 Van der Heijden, Rianne Erasmus MC Starting Grant: ’Pain Imaging’. 2024-2030 Oei, Edwin General Electric Healthcare: ‘RSNA QIBA MSK Profile Stage 3 and 4 Conformance Testing’. 2023-2025

184

Oei, Edwin Horizon 2020 EIC Accelerator: ‘AI algorithms in musculoskeletal radiography’. Main applicant: Radiobotics, Copenhagen, Denmark. 2020-2025 Oei, Edwin Independent Research Fund Denmark: ‘Investigating pathology and tissue biomarkers of Osgood Schlatter to enhance treatment of children with growthrelated pain’. Main applicant: M.Rathleff, University of Aalborg, Denmark. 2019-2024 Oei, Edwin NWO Zon-MW Open Competition: ‘Biomechanical precision diagnostics in osteoarthritis’. Main applicant: S. Bierma-Zeinstra (General Practice/Orthopedics). 2020-2025 Oei, Edwin NWO Zon-MW Goed Gebruik Geneesmiddelen: ‘Efficacy of antibiotic treatment for patients with chronic low back pain and Modic type I changes - randomised placebo controlled trial’. Main applicant: B.W. Koes (General Practice). 2023-2027 Oei, Edwin, Rianne van der Heijden, and Jukka Hirvasniemi TU Delft-Erasmus MC Convergence Flagship: ‘Healthy Joints’. Main applicant: S.M.A. Bierma-Zeinstra (General Practice), J. Harlaar (TU Delft). 2022-2027 Oei, Edwin ZonMw Gender en Gezondheid - Algemene onderzoeksronde: ‘The FOCUM human disease model for development of OA’. Main applicant: S.M.A. Bierma-Zeinstra (General Practice). 2019-2024 Oei, Edwin ZonMw Gender en Gezondheid - Algemene onderzoeksronde: ‘Diagnosis, prevalence and associated factors of osteoarthritis in adults with intellectual disabilities’. Main applicant: D.A.M. Maes-Festen (AVG/General Practice).

scientific report 2025 | ADVANCED MUSCULOSKELETAL IMAGING RESEARCH ERASMUS MC (ADMIRE)

2020-2026

Society for Magnetic Resonance in Medicine and Biology (ESMRMB), Marseille, France. Oct 2025.

Oei, Edwin National Health and Medical Research Council, Australia: ‘SUPER rehabilitation RCT for young people with old knees’. Main applicant: K. Crossley (La Trobe, Melbourne, Australia). 2018-2024 Oei, Edwin European Research Council (ERC) Advanced grant: ‘Biomechanical precision diagnostics in osteoarthritis Modelling trajectories and mechanisms of childhood hip dysplasia’. Main applicant: S.M.A. Bierma-Zeinstra (General Practice/Orthopedics). 2023-2028

Invited Lectures

Edwin Oei. ‘New imaging techniques: photon counting CT’. Nederlandse Vereniging voor Calcium- en Botstofwisseling (NVCB), Zeist, the Netherlands. Nov 2025. Edwin Oei. ‘Imaging of knee osteoarthritis’, ‘Photon counting CT: a novel tool to assess total knee arthroplasty’, and ‘RCT with Sham Control for Knee Osteoarthritis at Erasmus: Lessons Learned’. Three lectures during Vascular Embolotherapy and Nerve interventions Targeting Inflammation and pain (VENTI), Tokyo, Japan. Nov 2025.

Edwin Oei. ‘How to train radiologists’. Erasmus Photon Counting CT Meeting, Rotterdam, the Netherlands. Jan 2025.

David Hanff. ‘Game-changing ZTE Bone MR imaging for young athletes: a new standard without CT’. Dutch Conference of Sports and Exercise Medicine, Papendal, the Netherlands. Nov 2025.

Edwin Oei. ‘Old and new techniques to assess bone status’. European Congress of Radiology, Vienna, Austria. Feb-March 2025.

Rianne van der Heijden. ‘Imaging of the spine’. Dutch Society for Anesthesiologists Pain fellowship educational summit, Bunnik, the Netherlands. Nov 2025.

Wouter Schallig. ‘Imaging during gait: a novel perspective on osteoarthritis’. Dutch Association for Technical Medicine (NVvTG) conference, Utrecht, the Netherlands. March 2025.

Additional Personnel

Rianne van der Heijden. ‘Bridging the Gap between Technical Developments & Clinical Needs’. International Society for Magnetic Resonance in Medicine (ISMRM), Honolulu, USA. May 2025. Rianne van der Heijden. ‘The potential of FDG PET/MRI for the identification of the pain generator in chronic pain’. GE HealthCare SIGNA Masters PET/MRI Nordic meeting, Helsinki, Finland. April 2025. David Hanff. ‘Game-changing ZTE Bone MR imaging for young athletes: a new standard without CT’. Joint meeting of the Belgian and Dutch MSK Sections (BSR-NVvR), Mechelen, Belgium. May 2025. Edwin Oei. ‘Genicular artery embolization and Radiofrequency ablation for knee osteoarthritis: promise or sham?’. EULAR European Congress of Rheumatology, Barcelona, Spain. June 2025. Edwin Oei. ‘Photon counting CT for osteoarthritis’. International Workshop on Osteoarthritis Imaging (IWOAI), Cambridge, UK. July 2026. Edwin Oei. ‘Advanced imaging of osteoarthritis’, ‘ The roadmap of (MSK) imaging biomarkers’, and ‘Impingement syndromes of the lower extremity’. Three lectures during Visiting Professorship at the Department of Radiology, Mayo Clinic, Rochester, USA. Sept-Oct 2025. Edwin Oei.

‘AI

musculoskeletal

MRI’.

European

Galied Muradin, MD, PhD - Musculoskeletal radiologist Mariëlle Olsthoorn, MD - Musculoskeletal radiologist Michael Ananta, MD, PhD - Nuclear medicine physician Koen Willemsen, MD, PhD - Affiliated post-doc focusing on 3D imaging and printing Erin Macri, PhD - Affiliated assistant professor from Dept. of Orthopedics focusing on biomechanics Guillaume Renaud, PhD - Affiliated associate professor from TU Delft focusing on ultrasound of bone Eveline Molendijk, MD - Affiliated researcher from Dept. of General Practice

Visiting researchers and interns Netanja Harlianto, MSc - Student Medicine, University of Utrecht. Until mid-2025. Daily supervisor Jukka Hirvasniemi. Lucas Bronder, MSc - NIHES Research Master Health Sciences student, Nov 2023-Sep 2025. Daily supervisor Jukka Hirvasniemi. Yijie Fang, MD - Visiting radiologist/researcher, The Fifth Affiliated Hospital, Sun Yat-Sen University, China. Sep 2024-Sep 2025. Maryam Mofid, MD - Research Master Student Clinical

185

Epidemiology, Dec 2024-July 2025. Daily supervisor Mirthe Kamphuis. Joost van den Broek, BSc - Student Biomedical Engineering, TU Delft. 2024- July 2025. Daily supervisor Wouter Schallig. Mark Planken, BSc - Student Biomedical Engineering, TU Delft. 2024- September 2025. Daily supervisor Wouter Schallig.

Dorte Waasdorp, BSc - Student Biomedical Engineering, TU Delft. From Oct 2025. Daily supervisor Wouter Schallig. Sebastiaan Koninkx, BSc, Student Technical Medicine. MarJun 2025 and From Nov 2025. Daily supervisor Rianne van der Heijden. Michiel Aten, BSc, Student Technical Medicine. From Nov 2025. Daily supervisor Rianne van der Heijden.

Josefien van den Berg, BSc - Student Technical Medicine, TU Delft. Dec 2024-Sep 2025. Daily supervisors Marijn Mostert and Rianne van der Heijden. Lianne Damink, BSc - Student Biomedical Engineering, TU Delft. 2024- May 2025. Daily supervisor Wouter Schallig. Imme Spiekerman, BSc, Student Technical Medicine. From Sep-Dec 2025. Daily supervisor Rianne van der Heijden. Colin Middel, BSc - Student Human Movement Sciences, VU Amsterdam, From Sep 2025. Daily supervisor Wouter Schallig. Julia Lens, BSc - Student Human Movement Sciences, VU Amsterdam, From Sep 2025. Daily supervisor Wouter Schallig.

Post-doc

Wouter Schallig, PhD Combining imaging and biomechanics to unravel knee osteoarthritis In 2040, osteoarthritis is the most common disease In the Netherlands affecting 1.8 million people according to the RIVM. Forces affecting the knee play an important role in the development of osteoarthritis. This, so called, joint loading can be quantified with biomechanical measurements. However up until now these measures have not been sufficiently accurate to quantify the cartilage loading. My research focusses on developing the first lab In the Benelux that Is combining fluoroscopy Imaging with the traditional marker-based motion capture technology and subsequently applying it to answer clinically relevant questions. With this combination of techniques, we can accurately quantify bone movements, joint loading and cartilage strain and stresses. In the LoaD project we apply these measurements to patients with knee osteoarthritis during walking, running and cycling. We will also relate those loading parameters to the outcomes of PET/MRI scans to better understand the relation between loading and biology within osteoarthritis and eventually improve personalized treatment and prevention strategies.

186

scientific report 2025 | ADVANCED MUSCULOSKELETAL IMAGING RESEARCH ERASMUS MC (ADMIRE)

Assistant Professor

Rianne van der Heijden, MD, PhD Email r.a.vanderheijden@erasmusmc.nl

Rianne van der Heijden is an academic radiologist with dual expertise in Radiology and Nuclear Medicine, specializing in musculoskeletal imaging. As a physician–scientist, she advances translational (PET) MRI by bridging physics and clinical research. She obtained her PhD on the mechanisms of knee pain using advanced MRI, including the development of quantitative DCE-MRI perfusion pipelines. During her residency, she received the Young Investigator Award from the ESSR and contributed to international standardization efforts through ISMRM-OSIPI and the QIBA DCE-MRI initiative. After completing her residency, she was a Visiting Assistant Professor at the University of Wisconsin–Madison, focusing on translational PET/MRI, while also coleading a randomized clinical trial on FDG PET/MRI in chronic musculoskeletal pain at Erasmus MC. In 2023, she was selected as Junior Fellow of the ISMRM. In 2024, she was honored with a personal ZonMW VENI grant to further develop her research line in pain imaging.

Musculoskeletal pain Chronic musculoskeletal pain is a significant burden for both patients and society. Many patients receive suboptimal treatment or no treatment at all, largely because current anatomy-based imaging techniques cannot accurately identify the underlying causes of pain. Innovative hybrid molecular imaging approaches, such as 18F-fluorodeoxyglucose (FDG) and Manganese (Mn) PET-MRI, have emerged to address this gap. These tracers target the underlying pathophysiological changes associated with pain, such as low-grade inflammation and heightened neuronal excitation. Additionally, lowgrade inflammation can also be detected using perfusion imaging techniques, such as dynamic contrast-enhanced MRI. These advanced diagnostic imaging methods have demonstrated promise in identifying pain generators in an ongoing randomized controlled trial (RCT) involving patients with chronic low back and hip pain. Future research aims to explore other PET tracers that may offer greater specificity for distinct pain pathways, further advancing diagnostic capabilities in this field.

Technical advances My research bridges technological innovation and clinical applications. I am actively involved in advancing PET/MRI acquisition and postprocessing techniques as a member of the Hybrid Imaging and Theranostics Center of Excellence. This includes improving PET/MRI attenuation correction near metal, which is essential for accurate imaging in patients with implants, and developing automated tools for PET lesion detection in

Figure 1. FDG PET/MRI in a patient with chronic low back pain. MRI shows morphologic abnormalities at both the left and right facet joint, while the combination with FDG PET only shows the right facet is metabolic active as a sign of low-grade inflammation. Facet denervation followed, leading to pain relief.

patients with chronic pain to shorten reporting time. A new endeavor is dynamic PET scanning of NaF and FDG. Another key area is optimizing quantitative DCE-MRI analysis to facilitate clinical translation. As part of this effort, I have been a longstanding contributor to QIC, former QIBA, and OSIPI. I am a part of the Advanced Musculoskeletal Imaging Research group (ADMIRE), where I co-supervise multiple PhD students and master students. I am also appointed as Adjunct Clinical Assistant Professor at the Department of Radiology, University of Wisconsin -Madison, USA.

187

PhD Students Daniek van der Kaaij, MSc

David Hanff, MD, MSK radiologist

Advisors Edwin H.G. Oei, Annette van der Helm - van Mil & Pascal de Jong

Advisors Edwin Oei, Adam Weir, Rintje Agricola & Joshua Heerey Hip and groin pain in athletes In collaboration with the Department of Orthopedics and Sports Medicine and the La Trobe University, Melbourne, Australia. I investigate clinical and imaging findings in athletes with hip and groin pain. Main focus is on the pubic symphysis and the hip concerning FAI (femoro-acetabular imgipingement). Discovering new features of MR imaging technique called oZTEo giving CT like images on MRI.

Dixon MRI for clinically suspect arthralgia MRI-detected subclinical joint inflammation is a promising predictor of progression from clinically suspect arthralgia to rheumatoid arthritis. Our project aims to optimize the Dixon MRI sequence and enable automatic detection of subclinical Inflammation in the hands. This will Improve the feasibility of MRI for routine rheumatology practice.

Desiree de Vreede, MD, MSc

Marleen van den Heuvel, MD, MSc Advisors Marienke van Middelkoop, Edwin Oei, Jeroen Renkens & Sita BiermaZeinstra

Advisors Edwin Oei & Aad van der Lugt MRI of the hips in the Generation R study This research Is part of a population-based prospective cohort study following children from fetal life until adulthood (Generation R). A total of 2863 MRI scans of the pelvic and hip area were collected at 10.2 years (mean, range 8.6-12.9 years) We aim to identify genetic and environmental causes of normal and abnormal growth and development of the lumbar spine and hip.

Growing Up: lifestyle and jointh health Structural spinal abnormalities and shape variations might play a role in the development of back pain in children and later in life, and have been shown in children on MRI. For this project data of children aged 9 years old from the Generation R Study was used to assess the prevalence of structural spinal abnormalities and shape variations, and to investigate associations with weight status and physical activity level of the children.

Sanne Boeren, MD, MSc Advisors Annette van der Helm- van Mil, Edwin Oei & Pascal de Jong

Marijn Mostert, MSc Advisors Edwin Oei & Rianne van der Heijden

Short MRI scan for clinically suspect arthralgia Patients with Clinically Suspect Arthralgia are at risk for Rheumatoid Arthritis. MRI detected subclinical joint inflammation is an important predictor. A short modified Dixon sequence has been developed (in the LUMC) to reduce scan time and lower MRI-costs. We study its accuracy in the Rotterdam CSA cohort, the relation with ultrasound and we determine the cut-off for positivity in an “ATLAS-study”. This project takes place in the Medical Delta. Ultimately, we envision that scans will be processed with AI and that a rapid scan is followed by a rapid answer.

F-FDG PET/MRI for chronic pain Chronic lower back pain and hip pain have a very large burden on patients and the healthcare system. Accurate identification of sources of pain in these patients is not straightforward, and conventional imaging techniques are insufficient. The AMPHiBI Trial aims to evaluate the value of 18F-FDG PET/ MRI for identifying chronic pain generators in a randomized controlled clinical trial. In addition, I work on technically oriented projects in the field of PET/MRI and DCE-MRI.

188

scientific report 2025 | ADVANCED MUSCULOSKELETAL IMAGING RESEARCH ERASMUS MC (ADMIRE)

Jie Deng, MD, MSc

Rosemarijn van Paassen, MSc

Advisors Edwin Oei, Robert-Jan de Vos & Denise Eygendaal

Advisors Edwin Oei, Marienke van Middelkoop & Sita Bierma-Zeinstra

Advancing the management of Patellar tendinopathy I investigated the prognostic value of commonly Imaging measures including MRI and conventional ultrasound in predicting patellar tendinopathy treatment outcomes following exercise rehabilitation.

Semi-automatic segmentation of knee MRI The developed semi-automatic segmentation algorithm, consisting of a spatial and an appearance component, is accurate and can be applied to large MRI datasets. This method enables us to explore the association between shape variation, development of the knee and the presence of knee pain in children in the population-based Generation R Study.

Laura Kemmeren, MD, MSc Advisors Edwin Oei, Duncan Meuffels & Tom Piscaer

Niels Dur, MD, MSc Advisors Jaap Harlaar, Edwin Oei, Erin Macri & Rianne van der Heijden

Mapping the adolescent knee: developmental abnormalities and osteochondritis dissecans We investigate different facets of growth, development and abnormal development of the adolescent knee, with a special focus on juvenile osteochondritis dissecans. This project is in collaboration with the Department of Orthopedics and Sports Medicine and uses Generation R data.

The role of biomechanics in OA By combining high-fidelity biomechanical measurements with multi-scale modeling, we aim to clarify the role of disturbances in joint biomechanics in early stages of osteoarthritis (OA). By correlating these measurements with findings on advanced PET-MR imaging, we hope to identify potential imaging-biomarkers for early-OA, which will aid in developing preventative or disease-modifying treatment strategies in future research.

Ilaria Orsi, MSc Advisors Guillaume Renaud, Rianne van der Heijden & Eric Verschuur

Tijmen Alexander van Zadelhoff, MD, MSc

Characterization of bone marrow lesions In osteoarthritic knees with ultrasound and MRI

Advisors Edwin Oei, Rianne van der Heijden & Koen Bos

Knee osteoarthritis is commonly associated with bone marrow lesions, which are linked to pain and disease progression. Our study evaluates a novel ultrasound technique for detecting and measuring these lesions by comparison with MRI, while also exploring its ability to assess bone blood flow and improve understanding of lesion pathophysiology.

Genicular artery embolization for knee osteoarthritis Genicular artery embolization (GAE) is a minimally invasive treatment for knee osteoarthritis (KOA) with promising results from multiple cohort studies. This project is a 1 year RCT to determine the efficacy of GAE in symptomatic KOA patients resistant to conservative therapy. We showed that GAE does not offer more pain reduction than sham treatment. Advanced MRI¬ including dynamic contrast-enhance MRI is applied to assess tissue changes following embolization.

Elisa Maijoor, MSc Advisors Edwin Oei, Frank Huygen & Rianne van der Heijden F-FDG PET/MRI for facet joint related pain Treatment of facet-related low back pain is no longer reimbursed in the Netherlands, as the pain generator often cannot be pinpointed with physical examination or conventional imaging. The PERFACT trial evaluates whether 18-FDG PET/MRI can identify the pain generator to guide targeted treatment in patients with facet related low back pain.

189

Aad van der Lugt, MD, PhD full professor 190

INTERVENTIONAL RADIOLOGY RESEARCH

Top

Publications

2025

Chen A, K Tey, R Verhage, R Maan, K Pieterman. Percutaneous Biodegradable Stent Placement for Treatment of Benign Biliary Strictures: A Systematic Review. J Vasc Interv Radiol. 2025; 36:556-563. Korenblik R, [..], K Pieterman, [..], C van der Leij, RM van Dam, Dragon Collaborative study group. Safety and efficacy of combined portal and hepatic vein embolisation in patients with colorectal liver metastases (DRAGON1): a multicentre, singlearm clinical trial. Lancet Reg Health Eur. 2025; 53:101284.

Bakas JM, MA de Wolf et al. Venous Stents Placed for Postthrombotic Syndrome: The Role of Inflow Disease on Patency. J Vasc Interv Radiol. 2025; 36:446-455.

His research focuses primarily on imaging and treatment of acute stroke. In addition, he leads national research efforts within the Dutch acute stroke CONTRAST Consortium. a.vanderlugt@erasmusmc.nl

Context

nterventional radiology is a rapidly expanding field of image-guided procedures allowing to obtain tissue samples or treat various diseases through a small hole in the skin.

In the Erasmus MC, many interventional radiology procedures are being performed yearly with a focus on advanced oncological interventions, hepatobiliary interventions, vascular (arterial and venous) interventions, urological and neuro-interventions. The Erasmus MC is a center of expertise for complex neuro-interventions, vascular malformations, hepatobiliary diseases, and oncological interventions, accredited by the European society of interventional radiology (CIRSE IASIOS certification). The section of interventional radiology is actively participating in scientific research, aiming at improving current treatment strategies and development novel therapies in the fields mentioned above.

191

Research Projects: Objectives & Achievements Projects: • An Imaging-based decision support tool which mechanically characterizes venous thrombi to aid in personalized endovascular thrombectomy device selection (Mark de Wolf) •

Advanced hepatobiliary and mesenteric vascular interventions (Kay Pieterman)

•

Advanced algorithms for quantitative evaluation of cerebral DSA images after cerebral thrombectomy (Sandra Cornelissen)

•

Improving diagnostic accuracy and treatment value of chronic mesenteric ischemia patients (Eva Bocharewicz)

The HJ trial – biodegradable stent placement versus standard-of-care for treatment of posttransplant anastomotic biliary strictures Biliary complications are among the most frequently observed complications of liver transplant surgery and include postischemic strictures of the bile ducts and anastomotic bile duct strictures at the level over the hepaticojejunostomy. For anastomotic strictures, treatment approaches differ per center. Standardized treatment approaches are lacking with no comparative trials between the different treatment methods conducted thus far. Treatment trajectories often take several months, with multiple treatment sessions necessary to resolve the obstruction. We are currently in the final stage of writing the protocol of a pilot study (multicenter) to evaluate safety and efficacy of direct percutaneous biodegradable stent placement for anastomotic bile duct strictures after liver transplant or hepatobiliary surgery. When the results of the pilot are promising, we will continue with a multicenter randomized controlled trial, in which percutaneous biodegradable stent placement will be compared with current standard of care at the Erasmus MC (stricture dilatation using internalexternal biliary drainage tubes). We expect to submit the study proposal to the METC in the upcoming months. The project is conducted by dr. Kay Pieterman (study PI, Interventional radiology) in close collaboration with the department of gastro-enterology and hepatology (dr. Raoel Maan, co-principal investigator) and surgery (prof. Bas Groot Koerkamp).

Flush trial – effects of catheter flushing on cholangitis and hospitalization in patients with biliary drainage tubes Patients with biliary disease are frequently treated using percutaneous biliary drainage tubes to restore flow of bile from diseased bile ducts to the bowel. In this ongoing RCT we investigate whether patients treated with biliary drainage tubes have better outcomes when biliary drainage tubes are flushed 3 times a day using a syringe

of standard physiological saltwater solution. Patients are randomized between either flushing or no flushing of the drain. We are currently halfway with inclusion of patients towards the desired sample size. The study is conducted by Kay Pieterman (study PI in close collaboration with the department of gastro-enterology and hepatology and department of hepatobiliary surgery (physician asssistant Chulja Pek and Prof. Bas Groot Koerkamp).

Evaluation of a standardized treatment protocol for patients with acute mesenteric and portal vein thrombosis Patients in the region presenting with an acute mesenteric vein or portal vein thrombosis are preferentially being referred to the Erasmus Medical center as a treatment center for vascular hepatology. Together with hepatologists and surgeons, a protocol was introduced in the clinic in order to achieve a more standardized treatment approach for this patient group. With this standardized approach we hope to improve patient outcome. We are currently evaluating how introduction of this protocol altered treatment outcome by comparing patients treated according to the protocol with a historical cohort of patients treated before the protocol was introduced. The study Is conducted by Sophie Ciere, master student and future PhD HPB surgery under supervision of dr. Kay Pieterman (Interventional radiology), dr. Sarwa Darwish Murat (gastro-enterology and hepatology) and dr. Jeroen de Jonge (HPB surgery).

Percutaneous direct portal pressure measurement using pressure wires – a pilot study The gold standard for assessing portal hypertension is currently hepatic portal vein pressure gradient measurement, in which the portal pressure is derived indirectly by measuring the pressure increase when inflating a balloon occlusion catheter in the hepatic vein. Although this method is elegant, it is known that derived pressures do sometimes over- or underestimate actual portal pressure. In the proposed pilot by Dr. Raoel Maan (hepatologist, EMC) and dr. Kay Pieterman (Interventional radiologist), we will evaluate the feasibility, patient satisfaction and procedural time when performing a direct portal pressure measurement. For this method we will introduce a pressure wire - a microwire with a pressure sensor at the tip - in the portal system using ultrasound guidance. We aim to show that this method is easy, safe and reliable for invasive direct portal pressure measurement and will also explore potential applications of this approach during portal interventions such as the TIPS procedure. An NVGE gastrostart grant of 10.000 euro was awarded for the proposed pilot study. The pilot study is planned to start in 2026.

A-IMAGIO study The long-term objective of the Ablation-IMaging and Advanced Guidance for workflow optimization in Interventional Oncology (A-IMAGIO) project is to

192

scientific report 2025 | INTERVENTIONAL RADIOLOGY RESEARCH

develop a standardized, accessible, low-complex, highprecision, end-to-end solution for treatment planning, needle guidance, and treatment evaluation for thermal liver ablation of HCC and metastatic liver disease. We are currently in the final phase of METC and contract finalization before we will start including patients in the Erasmus MC. This multicenter trial is coordinated from Leiden University Medical Center (PI mark Brugmans, local PI Kay Pieterman).

Pressure study – intra-arterial pressure measurements during mesenteric artery stenting in chronic mesenteric ischemia. The Multicenter Pressure trial coordinated by Interventional Radiology Erasmus MC was successfully initiated at multiple study sites in 2025 with the first patients being included in the Erasmus MC. In this prospective multicenter trial that has currently started in the Erasmus MC and the majority of participating centers, we aim to perform intra-arterial pressure measurements during mesenteric artery stenting. The study will be coordinated by Eva Bocharewicz and Mathijs van Dongen (PhD’s on chronic mesenteric Ischemia (department of radiology and gastroenterology), and supervised by Dr. Kay Pieterman (study PI), Dr. Leemreis, prof. Bruno and Prof. Siersema (dept of gastro-enterology and hepatology). The aim of this study is to evaluate whether pressure measurements can predict patient symptom relief of mesenteric artery stenting in chronic mesenteric ischemia. We hypothesize that mesenteric artery pressure measurements across a stenosis can provide additional information about its hemodynamical significancy, because it is known that extensive collateral networks between the different mesenteric arteries exist, with possible compensation of flow through collateral circulation pathways. We hypothesize that pressure measurements across the stenosis can provide information on the sufficiency of the collateral circulation in maintaining mesenteric artery flow, thereby allowing us to predict which patient will benefit from stent placement. This is relevant, hence approximately 30% of the patients treated for suspected chronic mesenteric ischemia do not experience symptom relief after mesenteric artery stenting.

ALPrES²MA trial (Anastomotic Leakage Prevention by Endovascular Stenting of the Superior Mesenteric Artery) In this multicenter Dutch trial, we will investigate whether placing a stent in the middle intestinal artery prior to colon surgery can reduce dangerous bowel leaks (anastomotic leakage). For this study, Kay Pieterman will serve as an independent observer to assess the severity of a mesenteric artery stenosis and board member of the data safety and monitoring board.

Figure 1. Biodegradable biliary stent (Ella CS, s.r.o., Czech Republic) made of bioresorbable polymer (Polydioxanone), specifically designed for treatment of non-malignant biliary strictures.

Dragon PLC trial – portal vs. double vein embolization for future remnant growth in oncological liver resections The Multicenter Dragon Primary liver cancer RCT was initiated in the Erasmus MC to evaluate the effect of double vein embolization on liver growth prior to surgical liver resection. The multinational Dragon Primary Liver Cancer trial initiated by Maastricht Medical Center (MUMC) is currently open for inclusion of patients in the Erasmus MC. In this randomized controlled trial, we aim to study whether double vein embolization (portal and hepatic vein) results in more liver volume increase of normal liver tissue than portal embolization alone for patients with a unilateral primary liver tumor with insufficient liver remnant for resection. The study is funded by ZonMW and will run in different centers across Europe. The study is led by Maastricht Medical Center (PI Dr. Ronald van Dam) and CHU Liege, Belgium (Prof. Dr. Olivier Detry), with Sinead James PhD student and coordinating investor of the trial. Dr. Kay Pieterman is local PI of the study.

Expectations & Directions During 2026, novel techniques will be tested and introduced in the clinic and more collaborative research projects will be organized with a focus on hepatobiliary and vascular mesenteric interventions.

Additional Personnel Mathijs van Dongen – PhD on chronic mesenteric ischemia (Kay Pieterman co-promotor). Lui Hong – Research fellow assessing imaging biomarkers of treatment success in radiological oncology interventions Wenny Chen – Research fellow assessing imaging biomarkers of treatment success in radiological oncology interventions.

193

Post-docs

Sandra Cornelissen, MD, PhD Quantitative evaluation of cerebral DSA images after cerebral thrombectomy

During endovascular treatment of ischemic stroke the interventional radiologist uses fluoroscopic and digital subtraction angiography (DSA) images for navigation and for qualitative evaluation of treatment result. Algorithms for quantitative evaluation of DSA images are developed by the BIGR group under supervision of dr. Theo van Walsum in collaboration with Philips medical systems and dr. Cornelissen as clinical advisor. Dr. Cornelissen is co-promotor of PhD students Matthijs van der Sluijs and Frank te Nijenhuis. Recent projects involve automatic calculation of a more detailed TICI score than currently used in the clinic (autoTICI) and automatic perforation detection on DSA images. Dr. Cornelissen works as interventional radiologist and did an MSc in computer science in the past. She aims to build a bridge between technical research and clinical applications.

Kay Pieterman, MD, PhD Hepatobiliary and mesenteric vascular interventions

Kay Pieterman is an interventional radiologist at the department of radiology and nuclear medicine of the Erasmus MC. He has obtained his PhD during his bachelor’s and master’s degree in medicine on diffusion MRI imaging of the preterm brain, with different projects in the Erasmus MC and St. Thomas’ hospital London. After finalizing his PhD, Kay worked as a radiology resident in EMC and ETZ Tilburg, with subspecialty interventional radiology. After finalizing his residency, a clinical fellowship in interventional radiology was obtained at the EMC. Working as an interventional radiologist for 2 years now, Kay is currently conducting several research projects with a primary focus on advanced hepatobiliary and mesenteric Interventions. He is also co-promotor of Eva Bocharewicz, PhD candidate on chronic mesenteric Ischemia. Kay Is committee member of the Dutch Mesenteric Ischemia research group (DMIS) and committee member of the scientific board of the Dutch Society of Interventional radiology.

194

scientific report 2025 | INTERVENTIONAL RADIOLOGY RESEARCH

Mark de Wolf, MD, PhD Characterisation of deep venous thrombi & the treatment of Pelvic Venous Disorders

Deep venous thrombosis (DVT) is traditionally treated with anticoagulants, but modern therapies often involve mechanical clot removal using suction or stent-like devices. Research shows that a clot’s morphology and mechanical properties heavily influence the success of these procedures. Therefore, predicting these properties before intervention is crucial for selecting the optimal removal device. We aim to use MRI and CT imaging to predict clot mechanics and guide this clinical choice. Currently, the optimal management of pelvic venous disorder (PeVD) remains unstandardized. To address this, the interventional radiology and vascular surgery teams, in collaboration with the gynecological department of the Amsterdam UMC, are conducting two Delphi consensus studies: the first focuses on standardizing the diagnosis, while the second evaluates treatment strategies. Building on these insights, we will subsequently launch a clinical study investigating the efficacy of endovascular interventions for treating PeVD.

PhD Student Eva Bocharewicz, MD, MSc Advisors Marco Bruno, Peter Siersema, Desirée Leemreis - van Noord & Kay Pieterman

Diagnosis and treatment of chronic mesenteric ischemia My PhD focuses on diagnosing and treating patients with chronic mesenteric ischemia, through studies on intra-arterial pressure measurements, 13C-butyrate breath tests, CT-based stenosis segmentation, sex-based differences in presentation and outcomes, and cost-effectiveness of stent types-aiming to improve diagnostic accuracy, patient selection, and treatment value.

195

Pierluigi Ciet, MD, PhD associate professor

Daan Caudri, MD, PhD assistant professor 196

ADVANCED THORACIC IMAGING JOINT APPOINTMENT AT UNIVERSITY OF CAGLIARI (ITALY)

JOINT APPOINTMENT IN PEDIATRIC PULMONOLOGY

Pierluigi Ciet completed his radiology residency at University of Padova (Italy) in 2011. He obtained a PhD in chest magnetic resonance imaging (MRI) at Erasmus MC in 2016. During his PhD, he trained as thoracic radiologist at Beth Israel Deaconess Medical Center (BIDMC) and Boston Children’s Hospital (Harvard Medical School). After finalizing his PhD, he trained as pediatric radiologist at Erasmus MC Sophia Children’s hospital where since 2020 he works as staff pediatric and thoracic radiologist. Since 2020, he is also Associate professor of pediatrics and thoracic radiology at the University of Cagliari (Italy), and since 2025 also at the Erasmus MC – Sophia Children’s Hospital. Since 2021, he is chair of the thoracic section of the cardiothoracic imaging task force of the European Society of Pediatric Radiology (ESPR). Since 2022, he is also member of the research committee of ESPR and of the Artificial Intelligence and CT Taskforces. His main research line focuses on the development of chest MRI as new imaging tool for pediatric and adult pulmonary diseases. Since 2022, he coordinates the standardization of Photon Counting CT imaging for pediatric patient at Erasmus MC and within the ESPR.

Daan Caudri received a MSc in Clinical Epidemiology in 2006, trained at Erasmus University Rotterdam and Harvard Medical School Boston. In 2007, he obtained his MD at the Erasmus MC. He received an NWO-Top talent grant for his PhD research on childhood asthma. He completed his training in Pediatrics in 2015, followed by a fellowship in Pediatric Respiratory and Sleep Medicine at the Perth Children’s Hospital Western Australia from 2015-2018. He then joined the Pediatric Pulmonology and Allergology staff at the Sophia Children’s Hospital, and a joint appointment in Radiology since 2023. He served as member of the European Respiratory Society Guideline Working Group and from 2020-2024 was treasurer and secretary of the Netherlands Respiratory Society (NRS) and the Dutch Lung Congress (DLC). He is an active member of the European Board for Accreditation in Pneumology (EBAP) and member of the scientific board of the European Cystic Fibrosis Society Patient Registry. Since 2022, he is Director of LungAnalysis image analysis core laboratory for lung images. d.caudri@erasmusmc.nl

p.ciet@erasmusmc.nl

Context

Top Publications 2025 Raut P, Y Chen, A Taleb, M Bonte, ER Andrinopoulou, P Ciet, JP Charbonnier, CE Wainwright, HA Tiddens, D Caudri. Validation of an artificial intelligence-based automated PRAGMA and mucus plugging algorithm in pediatric cystic fibrosis. J Cyst Fibros 2025; 24:970-978. Lv Q, Y Chen, D Caudri, ER Andrinopoulou, W Kuo, JP Charbonnier, RJ Fleck, LR Soler, M Paoletti, F Vermeulen, G Morana, EY Lee, M de Bruijne, HA Tiddens, P Ciet. Normative values for lung, bronchial sizes, and bronchusartery ratios in chest CT scans: from infancy into young adulthood. Eur Radiol 2025; 35:4846-4860. Lovrenski J, M Raissaki, D Plut, E Alexopoulou, GB Görkem, HN Ozcan, J Geiger, D Gräfe, C Sileo, P Caro-Dominguez, P Ciet. ESR Essentials: imaging of common paediatric pulmonary diseases-practice recommendations by the European Society of Paediatric Radiology. Eur Radiol 2025; 35:5037-5052.

hest CT has superior sensitivity to detect early structural changes compared to lung function tests and can be performed from infancy. Previously the role of chest CT in the clinic was limited, mainly due to the lack of quantitative outcome measures and sensitive image analysis tools. LungAnalysis group of the Erasmus MC has been working on these issues in the past decade. Sensitive low-dose chest CT protocols have been developed and implemented in most of the European CF centers. PRAGMA-CF, PRAGMA-BPD, and BEST-CT have become among the most accepted CT outcome measures in several International randomized clinical trials. The historical limitation of using CT, namely the exposure to ionizing radiation has been substantially reduced by the latest Photon Counting Detector (PCD-CT) scanner. PCD-CT shows dose reduction up to 90% compared to standard CT scanner approaching the dose of few chest radiographs. While chest MRI can overcome this limitation by avoiding ionizing radiation, it is more complex than chest CT and also it does not match the new image quality provided by PCD-CT. Nevertheless, chest MRI is the optimal technique to obtain functional information related to ventilation, inflammation, perfusion, and structure (VIPS–MRI) in a single examination.

197

Research Projects: Objectives & Achievements Chest CT outcome measures Bronchiectasis and trapped air have been well validated as outcome measures in cystic fibrosis (CF), but traditional scoring systems were time-consuming and semi-quantitative. In close collaboration with scientists in Perth (PI: Stephen Stick), LungAnalysis has developed and validated sensitive alternative image analysis method (PRAGMA-CF) to score chest CT in children and adults. PRAGMA-CF has now been successfully used as primary outcome measure in two large studies (RCT’s in young children) and has proven to be substantially more sensitive to detect early CF lung changes than previously semi-quantitative scoring methods. In the last years several other disease specific manual scores have been developed and validated Including the PRAGMA-BDP for children with Bronchopulmonary Dysplasia and the BEST-CT score for non-CF bronchiectasis. This scores are being used as primary outcomes in investigator initiated studies as well as pharmaceutical randomized clinical trial. With the advent of AI capabilities we have also been working towards automated tools to quantify structural abnormalities. Ongoing collaborations with Thirona B.V. (Nijmegen), a Dutch company developing cutting-edge artificial intelligence strategies for quantitative lung imaging, have supported the development of a fully automated sensitive system to measure bronchus-artery (BA)-dimensions of all visible BA-pairs as well as automated mucus plug detection on chest CT. The algorithm has been integrated in Thirona certified software platform LungQTM which is installed at the virtual platform DRE at LungAnalysis Erasmus MC. Current work on the validation of both manual and automated tools are ongoing in a wide

variety of lung diseases such as CF (PhD student: Pranali Raut), non-CF bronchiectasis (PhD students: Federico Mollica), bronchopulmonary dysplasia - BPD (PhD student: Ieva Aliukonyte and Intern: Elzanne Speksnijder), post-infectious bronchiolitis obliterans - PiBO (Intern: Muhsen Al Sharad) severe asthma (PhD student: Ieva Aliukonyte), interstitial lung disease - ILD (PhD student: Cristina Cretu), chronic obstructive pulmonary disease COPD (PhD student: Tjeerd van der Veer), primary ciliary dyskinesia - PCD (PhD student: Federico Mollica), and immunodeficiencies (PhD student: Astrid van Stigt).

Standardized chest CT To make optimal use chest CT in clinical care & research and enable the use of automated scoring algorithm, standardized and optimized image acquisition is essential. Since 2020, LungAnalysis on behalf of the European CF Society (ECFS) has standardized 58 CF sites participating in the ECFS clinical trial network (CTN). Standardization has been carried out through an interactive website developed by LungAnalysis staff (https://lunganalysis.erasmusmc.nl/). LungAnalysis remains the official the CT-expert center of the ECFS-CTN, and still trains centers all over the world as part of new clinical studies using lung imaging outcome parameters. More recently, standardization efforts have shifted toward PCD-CT imaging, which enables a significant reduction in radiation exposure while offering new spectral imaging capabilities. This initiative is led by Dr. Pierluigi Ciet on behalf of the ESPR.

Chest CTs in modeling studies and patient registries In rare chest diseases, accurate imaging-based outcome measures are critical not only for clinical studies but also

Example of chest CT structural abnormalities scoring using both manual and automated image analysis tools. The top left inspiratory scan highlights: bronchiectasis in red, airway wall thickening in yellow, pink indicates atelectasis, and healthy lung tissue in green. The bottom left expiratory scan highlights: hypo-attenuation and/or air trapping in blue. The right image from inspiratory scan illustrates airway segmentation with mucus plugging shown in red.

198

scientific report 2025 | ADVANCED THORACIC IMAGING

for patient registries. In 2022, a large grant was awarded by the US CF Foundation (CFF) to analyze up to 10,000 CTs and incorporate imaging-based outcome measures in the European CF Society Patient Registry (ECFS-PR) (PI: Daan Caudri and PhD student: Pranali Raut). A similar project was conducted for the non-CF Bronchiectasis Registry (EMBARC) (PI: James Chalmers and PhD student: Angelina Pieters). The Dutch Asthma Fund awarded a grant to build the Dutch PiBO (Post Infectious Bronchiolitis Obliterans) registry, for which LungAnalysis will provide Image quantification.

MR Imaging of pediatric pulmonary disease LungAnalysis has been actively involved in chest MRI research since 2006, initially focusing on CF. The work has since expanded to include other respiratory conditions such as BPD and severe asthma. Through collaboration with the Department of Pediatric Surgery (PI: Marco Schnater), the chest MRI protocol has been successfully validated for assessing congenital lung malformations and implemented as a clinical follow-up procedure for children at the age of 8 years. In 2023, LungAnalysis obtained approval for the ADVANCER study, funded by the Horizon Pathfinder grant as part of the 3D Spiro MRI consortium (PI: Pierluigi Ciet and PhD student: Ieva Aliukonyte). The consortium’s primary objective is the development of innovative MRI methodologies for assessing pulmonary function and structure using lowfield MRI. The newly developed 3D Spiro-MRI technique will be validated against PCD-CT. It will also be applied in pediatric pulmonary imaging, including a large cohort of children with CF, BPD, and severe asthma. Advanced postprocessing tools will be used to compare spatial resolution and dose-related improvements with standard energyintegrating detector CT. Previous studies conducted by PhD students Bernadette Elders and Laurike Harlaar have enabled LungAnalysis to implement chest MRI protocols for evaluating upper airway conditions in both children and adults.

Axial 1 mm isotropic Photon Counting CT (PCCT) image and 1x1x5 mm T2-weighted PROPELLER MR image in patient with interstitial lung disease. Note how fibrotic tissue appears hyperintense on MR image compared to PCCT image (arrows).

Expectations & Directions LungAnalysis will continue to develop and validate automated chest CT scoring systems to quantify imagerelated biomarkers in close collaboration with image analysis companies such as Thirona B.V. (Nijmegen). LungAnalysis will also continue close collaboration with the Radiology Department of the Erasmus MC. In 2026, LungAnalysis will work towards automation of manual chest CT scoring systems currently performed in international clinical trials, such as PRAGMA-CF, PRAGMA-

MRI imaging for interstitial lung disease In 2025, considerable efforts were devoted to finalizing a novel chest MRI protocol for adult ILD within the framework of the M-ILD study (PI: Pierluigi Ciet and PhD student: Cristina Cretu). This comprehensive study integrates chest MRI, PET-MRI, and PCD-CT to achieve precise differentiation between fibrotic and inflammatory ILD. The study has almost completed baseline patient enrollment. The newly developed MR sequence enables acquisition of high image quality through deep learning (DL) denoising filters. Moreover, inflammation and fibrosis are quantified using T1 and T2 mapping techniques, Dynamic Contrast Enhancement and PET-MRI.

199

BPD, and BEST-CT. The roadmap and experience gained in the past decade with CF is currently applied to other lung diseases in both children and adults and may be applied to even more specific diseases, such as, PiBO and ILD. In 2026, LungAnalysis will complete the baseline and follow-up patient enrollment of the M-ILD study and finalize patient enrollment for the ADVANCER study. In parallel, LungAnalysis aims to develop a fully automated chest MRI scoring system for assessing conditions such as CF, BPD, and severe asthma, in collaboration with a research group in Australia (PI: Bowen Xin). LungAnalysis will also participate in multicenter studies investigating the use of chest MRI in immunocompromised children, as well as in pediatric and adult patients with diaphragm dysfunction.

Daan Caudri. ‘Early detection of lung damage in patients with PID by innovative CT analysis techniques’. WID Expert Meeting, Oisterwijk, the Netherlands. June 2025. Pierluigi Ciet. ‘Photon counting CT in children & Wise use of artificial intelligence (AI) in paediatric radiology: AI in thoracic imaging’. ECR, Vienna, Austria. Feb-March 2025. Pierluigi Ciet. ‘Application for Chest MRI in the Child’. RSNA, Chicago, USA. Nov-Dec 2025.

Additional Personnel Punitkumar Makani, PhD - Head of LungAnalysis Jorien van de Puttelaar, MSc - Project Manager

Funding Caudri, Daan, and Harm Tiddens American Cystic Fibrosis Foundation (CFF): ‘ENRICH the ECFS Patient Registry with Structural Lung data from Chest CT scans’. 2022-2026 Kinghorn, BreAnna, Daan Caudri, and Federico Mollica CD Foundation and GDMCC Early Career Investigator Award program: ‘Structural lung disease in children with PCD: utilizing an automated airway-artery method to detect disease progression’. 2023-2025 Caudri, Daan, Suzanne Terhegge, Pierluigi Ciet, and Punitkumar Makani Stichting Astma Bestrijding: ‘Developing and validating an AI-supported chest CT score to diagnose Post-infectious Bronchiolitis Obliterans (PiBO) in children’. 2024-2025 McNally, Paul, Harm Tiddens, and Daan Caudri CFF grant: ‘Real world outcomes with novel modifier therapy combinations in children with CF (ENHANCE study)’. 20232026

Merlijn Bonte, MSc - Lead Image Analyst Jana Milenković, MD - Radiology Resident Ahmad Taleb, MSc - Intern Beyza Yagmur Ikiz, MSc - Intern Elzanne Speksnijder, BSc - Intern Floortje Bregman, MSc - Intern Lin-Ting Wu, MSc - Intern Muhsen Al Sharad, MSc - Intern

PhD Students

Ciet, Pierluigi VENI Grant: ‘Magnetic Resonance Imaging in Interstitial Lung Diseases (M-ILD study)’. 2020-2025 Ciet, Pierluigi Horizon Pathfinder: ‘3D Spiro MRI project’. 2023-2025

Angelina van BeukeringPieters, MD Advisors Harm Tiddens, Menno van der Eerden, Pierluigi Ciet & Daan Caudri PhD Obtained: 18-12-2025

Invited Lectures Daan Caudri. ‘Bronchiectasis radiology masterclass’. EBROW, Paris, France. Feb 2025. Daan Caudri. ‘CT-scans for evaluation of cystic fibrosis lung disease, now and forever?’. ECFS, Milan, Italy. June 2025. Daan Caudri. ‘AI tools for early detection of bronchiectasis’. IPIC, Prague, Czech Republic. Nov 2025.

200

Image analysis and treatment of non-CF bronchiectasis Bronchiectasis Is characterized by permanently dilated airways, with chest CT scans crucial for its detection and quantification. Limited knowledge exists regarding the spectrum of structural lung changes in non-CF bronchiectasis patients. My PhD thesis explored structural lung changes, correlating quantitative CT findings with etiology and disease severity in the EMBARC registry.

scientific report 2025 | ADVANCED THORACIC IMAGING

Cristina Cretu, MSc

Federico Mollica, MD, MSc

Advisors Pierluigi Ciet, Thomas Koudstaal, Juan Hernandez-Tamames & Marlies Wijsenbeek - Lourens

Advisors Harm Tiddens, Pierluigi Ciet & Daan Caudri

Airway phenotyping of chronic lung diseases using chest CT scans

Magnetic resonance imaging for phenotyping and monitoring of interstitial lung disease: M-ILD study

Chronic lung diseases were characterized by progressive airway dilation, wall thickening, obstruction. My PhD research uses manual bronchiectasis scoring and artificial intelligence algorithms to analyze chest CT scans. It includes large studies in cystic fibrosis (CF), non-CF bronchiectasis, primary ciliary dyskinesia (PCD), and non-tuberculous mycobacterial (NTM) disease.

M-ILD is a prospective, longitudinal study that aims to develop an innovative MRI protocol for effective phenotyping, patient-tailored treatment, and monitoring of therapy response in patients with interstitial lung disease (ILD) through quantification of both fibrosis and inflammation.

Ieva Aliukonyte, MSc

Pranali Raut, MSc

Advisors Ricardo Budde, Harm Tiddens, Pierluigi Ciet & Daan Caudri

Advisors Daan Caudri & Harm Tiddens

Advanced imaging in chronic pediatric pulmonary disease: ADVANCER study

ENRICH: the ECFS patient registry with structural lung data from chest CT scans

ADVANCER study aims to validate a novel 3D Spiro MRI technique enabling lung function assessment using MRI. MRI findings in children with chronic pediatric pulmonary disease (CPPD) are compared with state-of-the-art photon-counting CT. The 3D Spiro MRI technique may enable combined functional and structural lung imaging in a single examination.

ENRICH is a retrospective, multi-center cohort study using real-world data from the European Cystic Fibrosis Society Patient Registry (ECFSPR). The project aims to enrich the registry with longitudinal structural lung data through automated scoring of 5,000 – 10,000 chest CT scans and to investigate the real-life impact of CFTR modulator therapy.

Tjeerd van der Veer, MD, MSc Advisors Joachim Aerts, Harm Tiddens, GertJan Braunstahl & Daan Caudri

Clinical and radiological studies in non-CF bronchiectasis, asthma, and COPD My thesis investigated non-CF bronchiectasis, obstructive lung diseases, and their overlap. It examined beclomethasone–formoterol for chronic cough and dupilumab in reducing exacerbations in ABPA. It also explored links between structural lung disease and phenotype, and automated CT analyses of bronchial dimensions, mucus plugging, and air trapping across different populations.

201

202

Focus Area

IMAGING IN HEALTH SCIENCES The ultimate aims are to better understand typical and atypical development over the life course, from childhood to old age; to identify health-related factors that can improve public health and inform healthcare policy; to unravel the etiology of illnesses; and to improve disease prediction and decision-making in clinical practice.

203

Meike W Vernooij, MD, PhD full professor 204

POPULATION IMAGING

Top Publications 2025

JOINT APPOINTMENT IN EPIDEMIOLOGY Meike Vernooij was trained in Radiology at Erasmus MC, followed by a clinical fellowship in Neuroradiology and Head & Neck Radiology. Meike combined her clinical work with a PhD in Neuroimaging/ Neuroepidemiology-completed cum laude in 2009-and has been working as a clinician-researcher in the internationally renowned Rotterdam Study since 2005. Her main interest is the study of brain changes on imaging in aging, and in particular imaging markers for stroke and dementia. Ultimate aim is to improve insight into etiology and prediction of these diseases. She was awarded the Lourens Penning Prize by the Radiological Society of the Netherlands and the Lucien Appel Prize in Neuroradiology by the European Society of Neuroradiology. She is principal investigator of Population Imaging of the Rotterdam Study and as such leading the Rotterdam Scan Study. In 2017, Meike was appointed as professor of Population Imaging, and in 2023 she was appointed as Medical Delta professor with an honorary appointment at TU Delft (Faculty of Applied Sciences). She has authored over 470 peer-reviewed publications and is currently leading a research group of 8 PhD students, 1 postdoc and 1 assistant professor.

Rosbergen MT, P van der Veere, JJ Claus, TE Evans, V Venkatraghavan, F Barkhof, AC van Harten, MA Ikram, van der WM Flier , MW Vernooij, FJ Wolters. Subcortical gray matter volumes and 5-year dementia risk in individuals with subjective cognitive decline or mild cognitive impairment: A multi-cohort analysis Alzheimers Dement. 2025; 21:e70413. Streiber AM, TC van den Beukel, I Vom Hofe, J Neitzel, MW Vernooij, D Bos, Vinke EJ. Arterial calcification in the heart-brain axis and cognitive performance over time. Alzheimers Dement. 2025; 21:e14374. Nguyen Ho PT, MW Vernooij, T Voortman, M Rodriguez-Ayllon, J Neitzel. Objective physical activity and Alzheimer’s disease burden in the population-based Rotterdam Study. Alzheimers Dement. 2025; 21:e70655.

m.vernooij@erasmusmc.nl

Context

nce patients present with symptoms, in many cases irreversible damage is already present. This is true for many diseases that are common in the population, like cardiovascular disease and dementia. If we want to understand how disease develops and which factors drive its development, we need to study disease in its earliest forms, when symptoms are not yet present. This is the area in which population-based studies operate. Studies that start out with presumed healthy individuals, assess potential disease determinants and follow participants for occurrence of disease. Imaging plays an important role in such epidemiological studies, by allowing us to non-invasively directly study the tissue at risk. Population Imaging, the large-scale acquisition of medical images in population-based cohorts, allows to investigate structural and functional changes that may indicate early disease, can be used to identify persons at risk of developing disease, or may aid in disease prediction. Our Population Imaging studies at Erasmus MC primarily take place within The Rotterdam Study, a prospective, population-based study aimed at investigating determinants of chronic and disabling diseases among nearly 18,000 persons aged 45 years and over. Since 2005, we are conducting brain MRI in all study participants (with repeat exams every 3-5 years). In selected subsamples, we have furthermore acquired NECT for vascular calcification, amyloid PET CT and 7T MRI for small vessel function.

205

Research Projects: Objectives & Achievements

of cognitive decline in a population setting and in memory clinic patients with subjective memory complaints.

Modelling normative brain aging This research line is focused on modelling normative brain aging using medical imaging and machinelearning technologies. Such normative models can help us understand where, when and how deviations from normal brain aging take place, with the ultimate aim to 1) better understand disease etiology, informing preventive strategies and 2) in a clinical setting, support earlier and more accurate diagnosis and monitoring of neurodegenerative or neurological disease. In this research line we have created normative data for MRI manifestations of cerebral small vessel disease, which can potentially aid clinicians in quantifying disease burden in individual patients and reveal specific patterns of abnormality (Figure 1). As part of the Medical Delta 2.0 project we are currently evaluating the implementation of the white matter hyperintensity volume percentile curves in the multidisciplinary meeting within the Alzheimer Center ErasmusMC. In this pilot study we assess the impact of the quantification of the burden of white matter hyperintensity volume on the etiological diagnosis of the memory clinic patients.

Legend Figure 2: Visualization of the three structural brain aging patterns identified in cognitively normal females in the Rotterdam Study. The colors at the different brain locations show Z-score changes across the different aging progression stages. One stage increase represents one additional Z-score change of one of the brain structure volumes. Of the 2521 females, 768 could not be subtyped (stage = 0).

Current projects (project lead: Eline Vinke): - Quantification of small vessel disease burden in TRACE-VCI using the normative small vessel disease data (Duygu Kilinc) - Disease progression modelling in normal ageing (Sterre de Jonge)

Etiology disease

Legend Figure 1: White matter hyperintensity percentile curves and the quantified white matter hyperintensity burden of a single patient, as presented in the pilot study in the multidisciplinary meeting of the Alzheimer Center ErasmusMC.

Another project within this research line focuses on identifying patterns of structural brain changes in normal aging, by applying disease progression modelling on the neuroimaging data from the population-based Rotterdam Study. We have identified 3 different brain aging patterns based on (regional) volumetric biomarkers of Alzheimers Disease, in both males and females (Figure 2). We are investigating whether we can use these patterns or the stage of progression within these patterns for prediction

and

prediction

Alzheimer’s

We have acquired one of the largest datasets of amyloid PET scans within a longitudinal European cohort, featuring over 600 participants from the Rotterdam Study. What sets this dataset apart is the extensive collection of genetic, health, behavioral, lifestyle, and brain pathology data. Our research focuses on two complementary goals: (1) investigating risk factors and underlying causes of Alzheimer’s disease and (2) developing prediction models to identify individuals at risk of developing Alzheimer’s disease. In terms of Alzheimer’s etiology, it is still not clear which role vascular brain pathology plays. One pathology may accelerate the other, or they may occur independently. So far, our evidence points towards the latter. We did not find an association of amyloid-beta plaques on PET with either arteriosclerosis on CT, white matter hyperintensities on 1.5T MRI or with small vessel functions on 7T MRI. We are currently investigating how vascular risk factors, such as diabetes and physical activity, relate to Alzheimer’s and/ or vascular disease, providing insights into how these risk factors mechanistically increase dementia risk. Finally, we

206

scientific report 2025 | POPULATION IMAGING

are developing prediction models for Alzheimer’s disease. To this end, we have enriched the PET sample with stateof-the-art blood-based biomarkers. We found that blood biomarkers perform extremely accurate in classifying amyloid-PET-positive versus -negative individuals (Fig. 2A). A larger increase of blood biomarkers levels was also associated with steeper cognitive decline and conversion to mild cognitive impairment (Fig. 2B).

Furthermore, within the Generation Next study, we collect brain imaging data in women who are trying to become pregnant. We scan our study subjects before pregnancy and after pregnancy (3 months and 1 year potpartum). The aim of this study is to better understand the remarkable impact that pregnancy has on the human brain, with a secondary focus on maternal mental health. Current projects: • Brain development, imaging trajectories and deviations in brain morphology in the pediatric population (Marjolein Dremmen) • Unravelling the neurobiology of anorexia nervosa (Katrien Bracke) • Decoding the maternal brain (Merel de Vries)

Discrimination amyloid-PET positive vs. negative

Expectations & Directions B

Risk of conversion from CU to MCI p-tau217 positive

Cognitive decline

p-tau217-negative

p-tau217 change rate

Legend Figure 3: Association between blood-based biomarkers and (A) Alzheimer’s disease on PET and (B) cognitive decline

Current projects (project lead: Julia Neitzel): Interplay between Alzheimer’s disease and arteriosclerosis and Alzheimer’s disease (Anna Streiber) - Interplay between Alzheimer’s and cerebral small vessel disease on cognitive decline (Anna van Houwelingen) - Prediction of Alzheimer’s disease and uncovering modifiable risk factors (Phuong Thuy Nguyen Ho)

Neurodevelopment in childhood and brain changes in pregnancy Within the population-based Generation R, we use population imaging methodologies to study brain development in children, from age 6 to young adulthood. We apply normative modelling to create reference data and apply these techniques for clinical translation (e.g. within the BRAVE study, to detect deviation from normality in girls with anorexia nervosa).

In the coming years, we expect that the amyloid PET data that were recently acquired will enable us to study the vascular and amyloid pathways towards dementia in depth, gaining more understanding in how these pathologies alone or in conjunction may lead to neurodegeneration and cognitive decline. Studies like the 7T small vessel disease study in collaboration with University Medical Centre Utrecht, will greatly add to this by allowing to directly image and measure microangiopathy, one of the most important contributors to both dementia and stroke. Furthermore, data-driven artificial intelligence research techniques (machine learning, deep learning) will bring the field of population imaging forward and improve our understanding of normal brain aging as well as neurodegenerative and neurological disease. Also, combining imaging with other high-dimensional data such as genomics, proteomics and metabolomics, is highly promising in unravelling pathways of disease and better understand disease pathophysiology. These features of arteriosclerosis may also be incorporated into existing risk prediction models for stroke and dementia, thereby supporting earlier identification and the development of preventive and therapeutic strategies. Another rapidly developing area is image post-processing using machine learning algorithms, which now allows much more detailed investigation of arteriosclerotic disease. In addition to conventional measures such as the Agatston score and volumetric assessment of calcification or plaque burden, it is becoming possible to assess shape characteristics, calculate vessel stress, and relate specific disease configurations to clinical outcomes. A further major achievement of this group was the completion of a large-scale follow-up imaging study with 10 to 15 years of follow-up, aimed at visualizing changes in vascular calcification across multiple arteries. These data are unique worldwide and have provided valuable insight into the natural course of arteriosclerosis.

207

Additional Personnel

Funding Vinke, Eline Dekker Postdoc Grant, Dutch Heart Foundation: ‘Personalized MRI-based Cerebral Small Vessel Disease (SVD) burden quantification, for more accurate diagnosis and prognosis’. 2024-2026

Anne-Sterre Schutter – Student Assistant MRI Ommoord Amanda Ros – Student Assistant MRI Ommoord Björn Schrijver – Student Assistant MRI Ommoord

Vinke, Eline Alzheimer Nederland Early Career grant: ‘Unraveling brain aging patterns predictive of AD or ADRD’. 2024-2025

Dagmar Mook – Student Assistant MRI Ommoord

Neitzel, Julia Erasmus MC Fellowship ‘Impact of Modifiable Factors on Dementia Risk across the Adult Lifespan’. 20242028

Fengli Bottema – Student Assistant MRI Ommoord

Vernooij, Meike and Esther Bron and Harro Seelaar ‘ZonMW TAP-Dementia: timely, accurate and personalized dementia diagnosis’. 2023-2027

Gaia Hermans – Student Assistant MRI Ommoord

Vernooij, Meike and Julia Neitzel Alzheimer’s Association Research Grant: ‘Amyloid pathology and vascular disease: a tale of two pathways’ - 2022-2025

Ismay van der Laar – Student Assistant MRI Ommoord

Vernooij, Meike, Frank Wolters and Arfan Ikram A public-private partnership receiving funding from ZonMW and Health Holland: ‘A Personalized Medicine Approach for Alzheimer’s Disease (ABOARD)’. 2021-2025

Esra Hemmelder – Student Assistant MRI Ommoord

Fenne Bosman – Student Assistant MRI Ommoord

Gunnar Zwart – Student Assistant MRI Ommoord

Issrae Affani – Student Assistant MRI Ommoord Lara Uitdewilligen – Student Assistant MRI Ommoord Lieke Bouvy – Student Assistant MRI Ommoord Levy Schimmel – Student Assistant MRI Ommoord – Team Leader

Invited lectures Meike Vernooij. ‘Bringing amyloid-treatment monitoring into daily clinical practice; invited presentation’. ECR, Vienna, Austria. Feb 2025. Meike Vernooij. ‘Imaging in Dementia’. lecture for the ESNR 2025 Basic Course in neuroradiology, Online. March 2025.

Lucas de Groot – Student Assistant MRI Ommoord Marije Timmermans – Student Assistant MRI Ommoord Mats Abspoel – Student Assistant MRI Ommoord Mehdi Badaoui – Student Assistant MRI Ommoord

Julia Neitzel. ‘Do commonly studied modifiable risk factors of dementia influence Alzheimer’s disease pathology? Findings from the Rotterdam Study’. EuroPAD Symposium, Barcelona, Spain. May 2025.

Ouidad Oujjit – Student Assistant MRI Ommoord

Meike Vernooij. ‘Update on the pathophysiology and neuroimaging of Amyloid Related Imaging Abnormalities (ARIA)’. University of Genoa, Genoa, Italy. May 2025.

Semih Cakmaktepe – Student Assistant MRI Ommoord

Paula Rijs Alonso – Student Assistant MRI Ommoord Team Leader

Ties Mulder – Student Assistant MRI Ommoord

Meike Vernooij. ‘Emerging Leader Reflections: Perspectives on challenges, future outlook, necessary resources, and potential solutions’. 16th Biennial International Symposium of the International Society for Strategic Studies in Radiology, Dublin, Ireland. Aug 2025. Meike Vernooij. ‘My top 5 tips for Dementia Imaging’. ESNR 2026 annual meeting, Istanbul, Turkey. Sept 2025. Meike Vernooij. ‘Hands-on session on ARIA imaging’. ESNR 2026 annual meeting, Istanbul, Turkey. Sept 2025.

208

scientific report 2025 | POPULATION IMAGING

Assistant Professor

Julia Neitzel, PhD Individualized dementia prevention Email j.neitzel@erasmusmc.nl

Julia Neitzel’s research interest is to understand why some people are more resilient against age-related cognitive decline and dementia – important, yet missing knowledge which could inform better prevention programs. For addressing this question, she combines her knowledge on cognition and brain imaging gained during her under- and postgraduate studies (MSc in Neuro-cognitive Psychology from Ludwig-Maximilians-Universität (LMU) Munich; PhD at the Neuroradiology of the Technical University Munich, Germany), with her postdoctoral work on genetic and modifiable risk factors of dementia (Institute for Stroke and Dementia Research at LMU; Depts of Radiology and Epidemiology at Erasmus MC). After her research visit at the Harvard School of Public Health in Boston, USA, she is excited to start her independent research group as an Erasmus Fellowship recipient. Prevention of chronic diseases is an ever-increasing forefront of medicine. Especially for dementia, once brain damage has occurred, treatment has been shown to be too late. My lab’s mission is to provide the knowledge for precision prevention strategies that promote brain health tailored to the demographic, genetic and clinical characteristics of each person. I support individualized dementia prevention through three key actions: (1) understanding the natural history of dementia, particularly mixed amyloid-vascular pathology, (2) detecting at-risk individuals, and (3) discovering modifiable risk factors for preventing Alzheimer’s disease pathology. Population brain imaging is the central component in my research, functioning as a powerful tool for early disease detection in asymptomatic individuals.

Identifying modifiable risk factors for preventing Alzheimer’s disease pathology The e4 allele of the Apolipoprotein E (APOE) gene is the major genetic risk factor for late-onset Alzheimer’s disease. About 25% of the Dutch population carries one copy, which increases their risk of Alzheimer’s by 2-3 times. To foster personalized prevention, we need to identify modifiable factors that accelerate Alzheimer’s pathology in high genetic risk groups. We have measured eight risk factors for dementia in participant from the Rotterdam Study. All participants underwent 18F-florbetaben PET about seven years later, on which we quantified their Alzheimer’s disease (amyloid-beta [Aβ] plaque) burden. In the high genetic risk group, we have found associations between more Aβ plaques and weak circadian rhythmicity, low cholesterol, and hypertension. Diabetes was associated with more Aβ plaques in both the low and high genetic risk group. We did not find a relationship between Aβ plaques and hearing loss, physical inactivity, obesity and smoking.

Figure 1. Associations between longitudinal plasma biomarker trajectories and (A) global and (B) regional Aβ-PET burden (Centiloids).

Blood-based biomarkers boost detection of Alzheimer’s disease pathology Biomarkers that detect Alzheimer’s disease in the blood could become cost-efficient tools to identify individuals at risk of developing dementia. Our prediction model incorporating demographic, genetic, and lifestyle factors demonstrated medium-to-high accuracy in classifying Aβ-PET status. (AUC=0.85). Adding plasma biomarkers significantly improved accuracy (AUC=0.94). With an Alzheimer Nederland grant, we studied longitudinal biomarker trajectories. We found that a more rapid increase in p-tau217 was the strongest determinant of Aβ-PET burden: a 1-SD change over time was associated with 145% higher Aβ plaques at follow-up, with smaller associations for GFAP (44%), Aβ42/20 (26%) and NfL (26%). (Fig 1.). Plasma p-tau217 was also the strongest determinant of cognitive decline and MCI conversion, performing comparably to Aβ-PET.

209

Post-doc

Eline Vinke, PhD Modelling brain aging

My research is at the intersection of neuroimaging, brain aging, epidemiology, and machine learning. I develop and apply data driven models to characterize normal and pathological brain aging using neuroimaging in combination with other biomarkers. A central premise of my work is that brain aging should be understood as a dynamic, multidimensional process rather than through isolated biomarkers. In 2023, I received the Early Career Grant from Alzheimer Nederland and the Postdoctoral Dekker Grant from the Hartstichting, which enabled me to establish my own research line focused on modeling healthy brain aging. My research explicitly investigates how multiple imaging derived and other biological markers change together over time, how they relate to one another, and how these relationships evolve throughout the aging process. This integrative perspective forms a key component of my scientific approach. By modeling joint trajectories and interdependencies between brain markers, I aim to capture heterogeneity in aging patterns and identify subtle deviations from normative aging. These normative, multimodal brain aging models can improve our understanding of disease etiology, inform preventive strategies, and support earlier and more accurate diagnosis and monitoring of neurological and neurodegenerative disease in clinical settings.

PhD students Merel de Vries, BSc Advisors

Katrien Bracké, MD, MSc

PhD Obtained 29-10-2025

Advisors Meike Vernooij, Gwen Dieleman & Tonya White

Meike Vernooij, Gwen Dieleman Advisors & Tonya White Meike Vernooij & Steven Kushner

Decoding the maternal brain

Unravelling the neurobiology of anorexia nervosa

My research focuses on how pregnancy changes the brain. In collaboration with the population-based Generation R Next study, we collect brain imaging data in women who are trying to become pregnant. We scan our study subjects before pregnancy and after pregnancy. The aim of this study is to better understand the remarkable impact that pregnancy has on the human brain, with a secondary focus on maternal mental health.

We aim to unravel the neurobiological underpinnings of anorexia nervosa (AN), using non-Invasive Imaging techniques. We use data of the BRAVE study, a clinical case-control sample of adolescents with firstonset AN compared to gender-, age- and educationmatched controls. A deeper understanding of the neurobiology may be a key element In developing novel therapeutic interventions.

210

scientific report 2025 | POPULATION IMAGING

Marjolein Dremmen, MD, MSc

Duygu Kilinc, MD, MSc Advisors Eline Vinke, Meike Vernooij & Geert Jan Biessels

Advisors Meike Vernooij & Tonya White

Improving diagnosis of SVD based on MRI scans: what is considered normal?

BRain development, Imaging trajectories and Deviations in brain morpholoGy in the pEdiatric population; BRIDGing thE gap

To bridge the translational gap between small vessel disease (SVD) imaging research and clinical practice, I will focus on creating age- and genderspecific normative data of the SVD markers using the Rotterdam Study data. Ultimately, I want to develop a tool that estimates the potential cognitive impact of SVD, based on lesion burden and distribution, in individual patients.

The goal is to construct ‘neuroimaging growth reference curves’ for total and regional brain volumes in children and adolescents. To be able to detect early deviations in brain development in patients groups; crucial for better understanding the neurobiology of these clinical conditions. To gain knowledge on variations in brain development and evaluate the consequences of incidental findings on brain imaging.

Phương Thủy Nguyễn Hồ, BSc

Mathijs Rosbergen, MSc

Advisors Julia Neitzel & Meike Vernooij

Advisors Frank Wolters, Meike Vernooij & Arfan Ikram

Modifiable risk factors for dementia prevention Unraveling MRI markers for dementia prediction

My research focuses on identifying which combination of modifiable factors are most effective for different age groups in relation to dementia. By examining how these factors influence imaging and cognitive outcomes, the goal is to better understand their role in dementia prevention alongside genetic and other unmodifiable risk factors.

When an individual is diagnosed with Alzheimer's disease, severe brain damage has already occurred. Treatment in an early stage could prevent brain damage and dementia. Therefore, my research focuses on unraveling new MRI predictors to improve existing dementia prediction models, so that we are able to identify individuals with high risk of dementia.

Anna van Houwelingen, MSc Advisors Meike Vernooij, Julia Neitzel & Harro Seelaar

Biomarkers for timely dementia diagnosis My research focuses on optimizing timely dementia diagnosis using plasma biomarkers and imaging markers like Aβ and tau-PET. By combining data from the Rotterdam Study and memory clinics, I investigate these biomarkers across both preclinical and clinical stages to improve diagnostic accuracy and guide patient management and treatment decisions.

211

Frank J Wolters, MD, PhD associate professor 212

CLINICAL EPIDEMIOLOGY OF NEUROLOGICAL DISEASE JOINT APPOINTMENT IN EPIDEMIOLOGY

Top Publications 2025

Frank Wolters obtained his MD at Utrecht University and practiced for a while in clinical neurology, before specialising in neuro-epidemiology with an MSc degree and PhD (cum laude) from Erasmus University Rotterdam. He completed research fellowships at the University of Oxford and the Harvard School of Public Health, applying expertise in research methodology notably to the prevention of cerebrovascular disease and dementia. His main interest lies with the understanding of dementia and stroke etiology, and improving the prognosis and management of subclinical, often incidental findings like covert brain infarction and asymptomatic carotid artery stenosis. Frank applies various modalities of brain imaging to further prevention and timely diagnosis of cognitive disorders, in clinical research as well as population-based studies. He leads the nationwide BIRD-NL consortium on dementia prevention in the Netherlands, and is a co-investigator on various national and international initiatives including the Cross-Cohort Collaboration, Dementia Risk Prediction Project, and the Heart-Brain Connection collaborative research group. Frank co-founded the working group for vascular cognitive impairment at the Alzheimer Centre, and since 2024, he is the principal investigator of Neuroepidemiology for the Rotterdam Study.

Yaqub A, JC Bis, S Frenzel, [...] & FJ Wolters. Clinical and imaging markers of cardiac function and brain health: a coordinated meta-analysis of communitybased studies. Neurology 2025; 104:e213421. Box CV, JA van Oortmerssen, MJ Tilly, MW Vernooij, MK Ikram, M Kavousi, FJ Wolters. Risk of cardiovascular disease with co-occurring covert brain infarction and unrecognized myocardial infarction. J Am Heart Ass. 2025; 14:e042282. Brück CC, SS Mooldijk, LM Kuiper, ML Sambou, S Licher, F Mattace-Raso, FJ Wolters. Time to nursing home admission and death in people with dementia: systematic review and meta-analysis. BMJ 2025; 388:e080636.

f.j.wolters@erasmusmc.nl

Context

imely diagnosis and prevention of neurological disease are key to reducing the burden of stroke, dementia, and other common neurological diseases in the population. Optimal prevention of neurological disease can only be achieved when clinical medicine and public health join forces. Through the study of population health in representative samples of the general population, often years before onset of disease symptoms, we learn about determinants of disease and subclinical changes that increase for example stroke or dementia risk. Similarly, risk stratification in these early stages allows for tailored diagnostic work-up and treatment, in order to optimize the benefitharm ratio of preventative interventions in primary care and hospital settings. Generating evidence from populationrepresentative data ensures applicability to the benefit of all patients in the community. Through the application of novel imethodology, we contribute to thorough and efficient acquisition and analyses of data for answering research questions that unravel disease etiology, and inform clinical practice and public health policy. As such, the strong bridge between clinical neuroradiology and population brain health is a reliable foundation for strengthening clinical guidelines as well as informing healthcare strategies.

213

Research Projects: Objectives & Achievements Bridging clinical and population science The Clinical Neuroepidemiology research group aims to improve the prevention of neurological disease, with a particular focus on stroke and dementia. We strive to translate findings from population-based research to the clinic, and vice versa, in order to unravel disease etiology and develop prognostic tools and treatments that are broadly applicable in medical practice. Key objectives are: 1. Identification of causes of neurodegenerative and neurovascular disease, disentangle underlying pathways, and quantify the contribution of these various causes to the clinical phenotype.

high lifetime risk of various neurological illnesses, along the 35-year course of the study, these observations have led for example to a diagnosis of dementia in over 2000 individuals, with equal numbers for stroke and 1000 cases with a diagnosis of migraine. The combination of detailed follow-up and population imaging creates a powerful combination for studying disease etiology and prognosis (Figure 1), Through incorporation of novel examinations in the core study protocol, we continuously look to enhance the phenotyping of participants with innovative techniques for a closer inspection of pathophysiological mechanisms in the early stages of these illnesses. Advances in research methods, like causal mediation analyses and target trial emulation, optimise data use for causal inference. We further apply population-based data to novel multistate models for determining the nationwide burden of disease and anticipated costs and effects of individual, often clinic-based, versus population-wide interventions.

2. Determine the relevance and optimal management of asymptomatic and subclinical pathology with respect to the onset and progression of neurovascular and neurodegenerative disease. Examples include asymptomatic carotid artery stenosis, covert brain infarction, and markers of cerebral amyloid angiopathy. 3. Develop and implement tools for timely diagnosis and prediction of neurological disease, with the aims of personalised prognosis, tailored treatment and follow-up, and targeted recruitment in clinical trials. Primary collaborators within the Clinical Neuroepidemiology research line are the department of Epidemiology (chair: professor Arfan Ikram), the Population Imaging group (lead: professor Meike Vernooij), the Biomedical Imaging Group Rotterdam, the Cardiovascular Imaging group (lead: dr. Daniel Bos), the Alzheimer Centre (departments of Neurology [lead: dr. Janne Papma] and Geriatric Medicine [chair: professor Francesco Mattace Raso), the department of Public Health (chair: professor Hester Lingsma), and the Generation R study (dr. Ryan Muetzel). Project: - Prediction tools for dementia (Mathijs Rosbergen)

The Rotterdam Study Epidemiological research on neurological disease within the Rotterdam Study currently entails the study of stroke, dementia, parkinsonism, migraine, epilepsy, traumatic brain injury, and polyneuropathy. The entire cohort of close to 18,000 individuals is monitored for the occurrence of these diseases through repeated in-person examinations and linkage to electronic health records. Owing to

Figure 1. Subcortical grey matter represents some of the oldest parts of our nervous system. In exploration for predictive markers, we analysed the prognostic value of volumes for progression to dementia in individuals with subjective cognitive complaints or mild cognitive impairment, in one population-based and two clinic-based cohorts. (Rosbergen et al. Alzheimers Dement 2025).

Alzheimer Centre Imaging Core In addition to the longstanding research cohort on frontotemporal dementia at Erasmus MC, 2024 has seen the birth of the Imaging Core of our Alzheimer Centre research at large. In close collaboration with BIGR and the Imaging Office, imaging data of over 800 patients have been processed, along with extraction of data from neuroradiology reports for research purposes. Requests for use of this data can be directed to the Alzheimer Centre executive board. The dataset will be expanded with a further 300 scans from the 2022-2025 period in the course of 2026, adding extra value to the research side of the Alzheimer Centre: research to improve care.

214

scientific report 2025 | CLINICAL EPIDEMIOLOGY OF NEUROLOGICAL DISEASE

Exemplary findings I: Arteriosclerosis across the heart-brain axis in vascular cognitive impairment and dementia Heart disease is an acknowledged contributor to stroke, and emerging evidence indicates heart failure and atrial fibrillation also predispose to dementia. Combining data from population-based and clinic-based studies like the Heart-Brain Connection Study, we unravel how cardiac dysfunction relates to brain pathology on MRI, and how this translates into clinical sequelae. An important area of interest is the overlap with Alzheimer’s disease pathology, which we can now measure in vivo using blood biomarkers. This has taught us for instance that biomarkers traditionally deemed important only to the central nervous system, appear to have systemic effects too that can clarify their role in vascular and other pathology (Figure 2). Moreover, we have shown that intracranial arteriosclerosis predisposes to dementia in the population-based Rotterdam Study. By currently taking these findings to clinical studies like the Alzheimer Centre Imaging Cohort, we translate these findings into clinically applicable tools for diagnosis and prediction that facilitate patient information and personalized care.

five older individuals, and increase the risk of subsequent cardiovascular disease and cognitive decline. Despite the large potential for secondary prevention, patients often go untreated, as optimal management is undetermined. Our group aggregated data from various cohorts to determine absolute risks for prognosis and trial design, and find determinants of high risk, which include infarct size and number, and concurrent unrecognized myocardial ischemia (Box et al. ESOC 2024). Ultimately, this can improve diagnostic and therapeutic management in clinical patients, like those visiting the Alzheimer Centre outpatient clinic or emergency department after head trauma. Project: - Prognosis and Clinical Management of Covert Brain Infarcts (Camiel Box)

Strengthening research methodology A reliable, open, and excellent academia is founded on rigorous standards for scientific quality. Through interaction with disciplines across Erasmus MC, the Clinical Neuroepidemiology group supports research on neurological disease and beyond, by offering support with expertise on study design, analyses, and causal inference. We consider methodological development an important part of PhD training, and are keen to support this whenever possible.

Expectations & Directions

Figure 2. In an interhemispheric comparison of people with unilateral carotid artery stenosis, embedded in the Rotterdam Study, we found that carotid artery stenosis causes cerebral microinfarcts and relates to lower parenchymal tissue volumes, possibly related to lower cerebral blood flow. This suggests cortical microinfarcts can be reflective of small vessel disease as well as large artery disease (Wolters et al. Cereb Circ Cogn Behav. 2025).

Exemplary findings II: Covert brain ischemia and stroke Covert brain infarcts are defined as ischemic lesions on imaging that were not preceded by (recognized) stroke symptoms. They are seen on routine brain MRI in one in

Population attributable fractions indicate that we can explain roughly 90% of the incidence of cardiovascular disease in the population, versus 45% for dementia. With our ongoing research, we aim to better understand the risk of dementia and move the percentage to the realm of cardiovascular disease. Similarly, on an individual patient level, we work to better understand which pathophysiological mechanisms are the principal contributors to the clinical phenotype (i.e. the etiologic fraction). For multifactorial diseases like cognitive impairment, this implies that we need to disentangle for example vascular contributions from those of Alzheimer’s disease, when there are multiple contributing causes present. The same applies to the identification of the most prominent risk factors in patients with stroke who have undetermined stroke etiology due to multiple ‘TOAST factors’ present. In terms of diagnostic strategies and risk stratification, we aim to further clinical management of covert brain infarcts by setting up a dedicated clinical service aimed at secondary prevention of vascular events and cognitive impairment. Accurate recognition and reporting of covert infarction on diagnostic imaging is crucial to serve this patient group well.

215

The applicability of new treatments against Alzheimer’s disease also requires better risk stratification, both to guide care referrals and identify individuals who may benefit most from intervention. This importantly includes understanding of the risk and impact of common adverse effects (i.e. microhaemorrhages and cerebral edema). To promote inclusive and generalisable science, we strive to maintain research that is applicable beyond tertiary patient populations. The Rotterdam Study is a key resource to this aim. We continuously look to incorporate innovative tools like spatial analysis of vascular pathology, functional vascular imaging, blood biomarkers, and digital lifestyle and cognitive assessments, to enhance our understanding of biological processes. Moreover, we explore the strengths and challenges of ‘big data’, for example through linkage of cohort data with Statistics Netherlands, Vektis, and IPCI, and look to strengthening ties with regional primary care, referral centres, and community old-age psychiatry services (e.g., Antes).

Funding Vernooij, Meike, and Frank Wolters, along with investigators from LUMC and TU Delft Medical Delta: ‘Applying advanced brain imaging for efficient dementia diagnosis and prediction’. 2024-2026 Wolters, Frank Alzheimer Nederland Biomedical Research Grant: ‘The APOE-ε2 paradox: role of APOE in lipid metabolism, vascular injury and amyloid deposition’. 2024-2026 Wolters, Frank, and Arfan Ikram, along with coinvestigators from 12 parties nationwide ZonMw National Dementia Strategy: ‘Modifiable factors for dementia prevention (BIRD-NL)’. 2023-2026

Frank Wolters. ‘Sex and gender differences in dementia: fact or artefact?’. Guest lecture Erasmus Summer Programme, Rotterdam, the Netherlands. Aug 2025. Frank Wolters. ‘Impact of asymptomatic vascular disease on cognitive health: implications for preventative care’. University of Wisconsin-Madison, Madison, USA. Oct 2025. Frank Wolters. ‘The joy of academic medical writing’. Fleming Fund Fellows seminar, Rotterdam, the Netherlands. Sep 2025. Frank Wolters. ‘Dementia risk oreduction in the Netherlands: a BIRD’s eye perspective’. Ministry of Health, Welfare and Sport, the Hague, the Netherlands. June 2025. Frank Wolters. ‘Prevention of dementia in individuals and populations: same-same but different?’ Amsterdam UMC, Amsterdam, the Netherlands. May 2025. Frank Wolters. ‘Prevention of dementia’. Alzheimercafé, Spijkenisse, the Netherlands. Feb 2025. Erik de Blois. ‘ Production and quality control of Ac-225 labelled radiopharmaceuticals’. NUMERI/NECSA, Pretoria, South Africa. March 2025. Erik de Blois. ‘ Ac-225 radiolabelling and GMP’. Institute of nuclear Technology, Online. Nov 2025.

Additional Personnel Amber Yaqub - project coordinator BIRD-NL Muhammed Sambou - PhD Student Amos Pomp - PhD Student Charlotte Menart - PhD Student

Wolters, Frank. Alzheimer’s Association Research Fellowship: ‘On generalizability and selection in dementia research: bridging academia and routine practice’. 20222025

Sanne van Kuijk – PhD Student Merel Huijg – PhD Student

Wolters, Frank. NWO Veni: ‘Prevention of stroke and dementia after silent brain infarction (SHINOBI)’. 20212025 Vernooij, Meike, Frank Wolters, and Arfan Ikram ZonMw and Health Holland: ‘A Personalized Medicine Approach for Alzheimer’s Disease (ABOARD)’. 2021-2025

Invited Lectures Frank Wolters. Epidemiology and risk factors of dementia. European Society of Radiology faculty course, online. Sept 2025.

216

scientific report 2025 | CLINICAL EPIDEMIOLOGY OF NEUROLOGICAL DISEASE

PhD Students Muhammed Sambou, MSc

Charlotte Menart, MD, MSc

Advisors Frank Wolters, Jeremy Labrecque & Arfan Ikram

Advisors Frank Wolters & Arfan Ikram

Modifiable lifestyle and environmental risk factors for dementia

Role of dual-task gait in neurodegenerative disease

Modifiable risk factors account for 45% of dementia incidence, but pathways are poorly understood and new risk factors needed. Using data from cohorts like the Rotterdam Study and Generation R / ORACLE, I apply causal inference approaches to clarify effects of environmental (e.g., pesticide exposure) and individuallevel risk factors (e.g., diet, antidiabetic therapy) on brain health and dementia risk.

Gait is the resultant of a complex interplay between sensory and motor function, and a such different gait domains can serve as markers of brain health and risk of neurodegenerative disease. I explore the added value of dual-task gait as a proxy of cognitive reserve, through investigating the within-person change between regular gait and gait whilst doing arithmic in the preclinical phase of dementia and Parkinson’s disease.

Jacqueline Claus, MD, MSC

Amos Pomp, BSc

Advisors Arfan Ikram, Meike Vernooij & Frank Wolters

Advisors Frank Wolters & Arfan Ikram

Understanding effects of APOE e2 and e4 in vascular injury and Alzheimer disease

PhD Obtained: 15-04-2025

The APOE gene is the most important genetic risk factor for dementia. Whilst most research is aimed at the risk increasing e4 allele, study of the protective e2 variant could identify neuroprotective strategies. Through study of lipids on vascular pathology and Alzheimer’s disease, including vascular and brain imaging, I strive to better understand the long-term effects of apolipoproteins on brain health.

Risk prediction of dementia: from theory to clinical practice Identification of modifiable risk factors for dementia, with a focus on early recognition and treatment of stroke and estimation of post-stroke dementia risk. Identifying those eligible for disease-modifying treatments against ß-amyloid. Izmprovement and implementation of the dementia prediction tool in a clinical setting.

Camiel Box, MD, MSc Advisors Frank Wolters, Meike Vernooij & Kamran Ikram

Prognosis and Clinical Management of Covert Brain Infarcts One in four elderly individuals are at high risk of stroke and dementia due to prior brain infarction that occurred unnoticed. In the Rotterdam Study we gather evidence on their prognosis and best clinical management in order to create clear guidelines to help clinicians deal with these covert brain infarcts.

217

Daniel Bos, MD, PhD associate professor 218

IMAGING OF ARTERIOSCLEROSIS: FROM POPULATION TO CLINICAL PRACTICE Top Publications 2025

JOINT APPOINTMENT IN EPIDEMIOLOGY Daniel Bos was trained as a medical doctor and epidemiologist at Erasmus MC. He obtained his PhD in the field of vascular imaging and vascular epidemiology in 2013. The main focus of his work is on the elucidation of the etiology and pathophysiology of arteriosclerosis, and the clarification of the contribution of arteriosclerosis to clinical neurovascular and neurodegenerative diseases. For his research, he leverages the strengths of populationbased studies and combines this with clinical patient studies to be able to make a direct clinical impact with his work. Over the years, he has been awarded the Best Scientific Paper Award by the Radiological Society of the Netherlands and the Lourens Penning Prize by the Radiological Society of the Netherlands. He is currently the deputy head of department of the department of Epidemiology, and holds a position as Adjunct Associate Professor in Clinical Epidemiology at the Harvard School of Public Health and a position as Professor in Neurosciences at KU Leuven. Daniel has authored over 150 peer-reviewed publications.

van der Bie J, BP Berghout, RP Budde, J Gutierrez, M van Straten, D Bos. Intracranial Calcification Detection; a Comparison Between Micro-CT, Conventional CT and Ultra-high-resolution PhotonCounting Detector CT. Acad Radiol. 2025; 32:35993607. Zuo L, M Kavousi, H Mo, MW Vernooij, D Bos. Evolution of Subclinical Carotid Atherosclerotic Plaque Composition Using Serial MRI in the Rotterdam Study. Radiology. 2025; 315:e242248. van der Bie J, BP Berghout, RP Budde, J Gutierrez, M van Straten, D Bos. Intracranial Calcification Detection: A Comparison Between Micro-CT, Conventional CT and Ultra-high-resolution PhotonCounting Detector CT. Acad Radiol. 2025; 32:35993607.

d.bos @erasmusmc.nl

Context

rteriosclerosis is a highly frequent vascular disease causing stiffening of arteries throughout the whole arterial system. Major clinical consequences of arteriosclerosis include myocardial infarction and stroke, which are top causes of morbidity and mortality in middle-aged and elderly persons worldwide. Due to the aging of the population, the global burden of arteriosclerosis, and thereby of these clinical conditions, will continue to rise in the coming decades. The research in “Imaging of Arteriosclerosis: from Population to Clinical Practice” is focused on the elucidation of the etiology and pathophysiology of arteriosclerosis, and on the clarification of the contribution of arteriosclerosis to clinical neurovascular and neurodegenerative diseases, with a strong emphasis on the use of state-of-the-art medical imaging. Within this unique research-field on the interface of arteriosclerosis, neurovascular, and neurodegenerative diseases, the research group specifically aims at obtaining relevant imaging-based biomarkers for improved understanding of arteriosclerosis and improved prediction of abovementioned diseases, and at identifying potential targets for intervention strategies.

219

Research Projects: Objectives & Achievements Intracranial Arteriosclerosis This is the major line of research within the group and focuses on the causes and consequences of intracranial arteriosclerosis. Our group has performed important pioneering work on the identification and quantification of calcification in the intracranial carotid arteries and vertebrobasilar system. Using non-enhanced computed tomography in participants of the population-based Rotterdam Study, we demonstrated that the presence and amount of intracranial carotid artery calcification is the strongest risk factor for a first-ever stroke, and that calcification at this location also contributes to the development of dementia, including Alzheimer’s disease. This work continues to have considerable global impact and has expanded our research towards clinical studies on the role of intracranial arteriosclerosis in stroke patients. In these patients, the burden of intracranial arteriosclerosis at hospital admission directly influences prognosis after stroke treatment. It has also led to valuable international collaborations on its influence on endovascular treatment, and its importance is increasingly being recognized in European Stroke Organization guidelines. Previous work in pathology and imaging revealed two morphological patterns of intracranial arteriosclerotic calcification: intimal, atherosclerotic calcification and circular calcification of the internal elastic lamina (IEL). Our group showed that these patterns contribute differently to stroke and influence stroke treatment, likely because of their divergent effects on arterial wall structure and the resulting mechanical stresses within the artery. Building on this, our research has expanded the role of these calcifications in neurodegenerative brain disease, including dementia. Initial findings suggest that circular IEL calcification in particular may contribute to neurodegenerative changes. In the coming years, we will investigate the mechanisms underlying this association by integrating information from multiple imaging modalities, including CT, PET, and MRI.

Advanced Imaging of Arteriosclerosis This line of research focuses on advanced imaging of arteriosclerosis, with particular emphasis on its most common subtype, atherosclerosis. Atherosclerosis is characterized by thickening of the intimal layer of the arterial wall due to lipid accumulation, neovascularization, and calcification. Especially at the carotid artery bifurcation, it substantially increases the risk of ischemic stroke. Research within this line is aimed at the optimal identification and evaluation of carotid bifurcation atherosclerosis. Using advanced state-of-the-art imaging

modalities, including CT and MRI, the group seeks to better understand the development of this disease and the value of plaque composition and morphology in predicting neurovascular and neurodegenerative diseases. Recent work from the group has shown that intraplaque hemorrhage in carotid atherosclerosis is the single most important risk factor for both first-ever stroke and stroke recurrence. In addition, our research demonstrated substantial differences in plaque composition between men and women, which remains an important topic for future study. An important recent focus within this line of research is the use of photon-counting CT (PCCT) for imaging arteriosclerosis. In particular, the resolution of conventional CT is too limited for imaging of cerebral vessels, and PCCT may offer an important new modality in this field. Current studies within the group compare PCCT with conventional CT for calcification detection, calcification quantification, and image quality, mainly using autopsy specimens (Figure 1).

Figure 1. Scanned cassette with zoomed-in samples. First column micro-CT, second column energy-integrating detector CT and last column photon-counting detector CT. Higher intensities are found with PCD-CT due to less partial volume effect than EID-CT due to the thinner slices of PCD-CT (row two and four). Smaller calcifications (third row) are more troublesome to detect by both scanners.

Optimization of Cardiovascular Image Analysis This research line revolves around the early detection of vascular disease, by maximally leveraging existing medical imaging examinations. By using novel methods for imageanalysis (e.g. machine-learning algorithms), we aim to

220

scientific report 2025 | IMAGING OF ARTERIOSCLEROSIS

obtain more sensitive and more accurate imaging markers for vascular disease. My group performed important pioneering work on the quantification of epicardial fat as emerging markers of vascular risk. Over the last years, this novel marker of vascular risk has gained rapid attention, which has led to a strong collaboration on its role in hearttransplantation patients with the Dept. of Cardiology and the Dept. of Pulmonology. This collaboration was further extended with additional measurements of imaging-based markers of vascular risk (liver density) and overall-health status (bone density), which can all be readily assessed using conventional CT (Figure 2).

These features of arteriosclerosis may also be incorporated into existing risk prediction models for stroke and dementia, thereby supporting earlier identification and the development of preventive and therapeutic strategies. Another rapidly developing area is image post-processing using machine learning algorithms, which now allows much more detailed investigation of arteriosclerotic disease. In addition to conventional measures such as the Agatston score and volumetric assessment of calcification or plaque burden, it is becoming possible to assess shape characteristics, calculate vessel stress, and relate specific disease configurations to clinical outcomes. A further major achievement of this group was the completion of a large-scale follow-up imaging study with 10 to 15 years of follow-up, aimed at visualizing changes in vascular calcification across multiple arteries. These data are unique worldwide and have provided valuable insight into the natural course of arteriosclerosis.

Invited lectures Daniel Bos. ‘The changing landscape of stroke: epidemiology and risk factors’. Fil Rouge of Atherosclerosis Conference, Cagliari, Italy. Oct 2025. Daniel Bos. ‘Intracranial Arteriosclerosis: A major risk factor for cognitive decline and dementia?’. VASCOG 2025, Southampton, UK. Sept 2025.

Figure 2. Structures that can be readily Identified on non-contrast cardiac CT are highlighted with different colors.

Expectations & Directions Research into arteriosclerosis is evolving rapidly, and major developments are expected at multiple levels in which this group will play an important role. Through the group’s work, intracranial arteriosclerosis is increasingly being recognized as an important risk factor for stroke and may also play a key role in the development of dementia. To advance these insights, detailed visualization of arteriosclerotic disease in the cerebral vessels using combinations of advanced imaging modalities will be essential. This includes not only the large afferent arteries, such as the carotid and vertebral arteries, but particularly the smaller cerebral arteries. An important focus will be the distinction between ‘pure’ atherosclerotic changes, such as intimal thickening and plaque formation, and arterial hardening or stiffening caused by calcification of the medial layer. Using population-based data, etiological differences between these forms of arteriosclerosis can then be identified.

Daniel Bos. ‘Intracranial Arteriosclerosis: Causes, Consequences, and Clinical Importance’. Seminar at Terumo Neuro, Los Angeles, USA. July 2025. Daniel Bos. ‘Carotid Plaque-RADS: Novel tool for Stroke Risk Classification?’. Nederlandse Vaatdagen, Noordwijkerhout, the Netherlands. April 2025. Daniel Bos. ‘Targeting Brain Arteries to Combat Dementia’. Lay talk for the Cure Alzheimer’s Fund US, online. April 2025. Daniel Bos. ‘Methodological Considerations when setting up an imaging biobank’. European Conference of Radiology, Vienna, Austria. Feb 2025. Daniel Bos. ‘Atherosclerosis and Dementia’. Public reading, Alzheimer Cafe Ridderkerk - Alzheimer Nederland, Ridderkerk, the Netherlands. Jan 2025.

221

Additional Personnel

Luca Bontempi – PhD student Department of Cardiology (co-promotor: Daniel Bos)

Céline van der Braak - PhD student Department of Radiology & Nuclear Medicine (co-promotor: Daniel Bos) (PI: Ivo Schoots)

Kostas Stoitsas - Postdoc Department of Epidemiology Hyunho Mo - Postdoc Department of Radiology & Nuclear Medicine (PI: Esther Bron)

Cevdet Acarsoy - graduated in 2025 - PhD student Department of Epidemiology (co-promotor: Daniel Bos)

PhD Students

Brian Berghout - PhD student Department of Epidemiology (co-promotor: Daniel Bos)

Aikaterini Tziotziou, MSc

Judith van der Bie - PhD student Department of Radiology & Nuclear Medicine (co-promotor: Daniel Bos) (PI: Ricardo Budde)

Advisors Ali Cagdas Akyildiz, Daniel Bos, Ton van der Steen & Aad van der Lugt

Luoshiyuan Zuo - PhD student Department of Epidemiology (co-promotor: Daniel Bos)

PhD Obtained: 26-08-2025

Mitra Nekouei - PhD student Department of Epidemiology (co-promotor: Daniel Bos)

Mechanical wall stress is associated with atherosclerosis development Atherosclerotic plaque onset and progression are known to be affected by local biomechanical factors. We investigated the association of mechanical wall stress (MWS) and wall shear stress (WSS) to atherosclerosis development in human coronary (IMPACT Study) and carotid (PARISK Study) arteries.

Peter van Hulst - PhD student Department of Neurology (co-promotor: Daniel Bos) Xi Li - PhD student Department of Public Health (copromotor: Daniel Bos) Sanne van Kuijk - PhD student Department of Epidemiology (co-promotor: Daniel Bos) Partha Haldar - PhD student Department of Epidemiology (co-promotor: Daniel Bos)

Assistent Professor

MAARTEN LEENING, MD, PhD Preventive Cardiology Email m.leening@erasmusmc.nl JOINT APPOINTMENT IN CARDIOLOGY AND EPIDEMIOLOGY Maarten Leening obtained his medical degree (cum laude) and MSc in clinical epidemiology at the Erasmus MC. He received his PhD (cum laude) at the Erasmus MC (Depts of Cardiology and Epidemiology). In 2014-2015, he worked as a post-doc at the Division of Preventive Medicine at the Harvard Brigham and Women’s Hospital (Nancy Cook and Paul Ridker). Subsequently, he trained as a cardiologist with a focus on non-invasive imaging (EACVI certified in cardiac CT and echocardiography), CVD prevention (EAPC certified), and lipids/atherosclerosis (EAS certified). By the end of 2023, Maarten Leening joined the vascular imaging research group (Daniel Bos). Atherosclerotic cardiovascular disease (ASCVD) events are preventable in the vast majority of individuals. My research aims at early life identification and treatment of cardiovascular risk factors and imaging of subclinical atherosclerosis in presymptomatic individuals. Specific interests Includes the biology and treatment of atherogenic lipids and chronic inflammation. On the side of imaging, areas of interest include non-invasive assessment of coronary atherosclerosis and aortic valve disease, and the role of automated software and AI in cardiovascular imaging.

222

scientific report 2025 | IMAGING OF ARTERIOSCLEROSIS

Xi Li, MSc

Anna Streiber, MSc

Advisors Hester Lingsma, Bob Roozenbeek, Theo van Walsum & Daniel Bos

Advisors Meike Vernooij, Julia Neitzel & Daniel Bos

In-hospital clinical decision support and treatment optimization for stroke

Disentangling the Role of Arteriosclerosis in Alzheimer’s Disease

This project aims to develop and validate traditional and AI models using pre- and peri-procedural clinical and imaging variables to estimate stroke patient outcomes after treatment. Additionally, we aim to create models for individualized prediction of treatment effect to optimize patient selection for endovascular treatment.

The etiology underlying Alzheimer’s disease is multifactorial and may be associated with cerebrovascular changes. Within the Rotterdam study, we investigate the relationship between arteriosclerosis and various markers for Alzheimer’s disease. These markers are measured in blood plasma, on (f)MRI, or PET-CT.

Ching Khan, MD

Judith van der Bie, MSc

Advisors

Aad van der Lugt, Daniel Bos & Bob Roozenbeek

Ricardo Budde, Daniel Bos & Marcel van Straten

The use of imaging to unravel the variability in outcomes after endovascular thrombectomy for ischemic stroke

Clinical applications of photon-counting computed tomography Photon-counting CT (PCCT) is a novel imaging technique which enables higher spatial resolution and more advance spectral imaging compared to conventional CT. During my PhD I will investigate the clinical impact of PCCT on cardiovascular- and neuroimaging. For example, the assessment of coronary in-stent restenosis with PCCT in patients with recurrent chest pain with invasive angiography as reference, to potentially minimalize number of patients undergoing this intervention.

The aim of my research is to find out through different imaging modalities why some ischemic stroke patients have worse functional outcomes than others when treated with endovascular thrombectomy. Ultimately, my goal is to identify ischemic stroke patients that may or may not benefit from thrombectomy treatment in clinical practice.

Luca Bontempi, MSc

Mohammad Nabil Loddin, MSc

Advisors Frank Gijsen, Behrooz Fereidoonnezhad & Daniel Bos

Advisors Daniel Bos, Kamran Ikram, Ali Akyildiz & Jolanda Wentzel

Enhancing Decision-Making in Endovascular Thrombectomy for Acute Ischemic Stroke

Carotid vascular disease and Brain Health

The overarching aim of my research is to support clinicians in the decision-making process for endovascular thrombectomy (EVT) in ischemic stroke patients by optimizing choices related to device selection, positioning, and treatment methods. Through the development and application of surrogate modeling techniques, the research seeks to enhance the efficacy of EVT, ultimately improving treatment outcomes for patients with acute ischemic stroke.

Investigating the role of carotid wall shear stress (WSS) on cerebrovascular outcomes. The project uses 3D reconstructed MRI data with a semi-automated pipeline for WSS in The Rotterdam Study, aiming to unravel the biomechanical processes within the carotid arteries that contribitue to the development of neurovascular disease and events.

223

Jacob J Visser, MD, PhD assistant professor 224

VALUE-BASED IMAGING

Top Publications 2025

Jacob Visser is a musculoskeletal radiologist, health economist and epidemiologist. As of July 2020, he is appointed as assistant professor in value-based imaging and working on issues such as artificial intelligence, structured reporting, decision support software, and integrated diagnostics. Furthermore, he is involved in setting up further collaboration with radiologists at the Radiology department of the Massachusetts General Hospital (MGH) in Boston, MA, USA. In addition, he is a member of the Value-based Radiology Committee of the ESR and the RSNA Working Group for Common Data Elements. He has served in the ESR eHealth and Informatics Subcommittee. As of January 1, 2020, he was appointed as Chief Medical Information Officer at the Erasmus MC.

Senkaiahliyan S, AS Detsky, J Ma, PR Lawler, J Paramasamy, JJ Visser. The Economic, Political, and Societal Consequences of Disrupting Health Care Delivery with Artificial Intelligence. J Gen Intern Med. 2025; 40:3473-3476. Paramasamy J, S Mandal, M Blomjous, T Mulders, D Bos, JGJV Aerts, P Vanapalli, V Challa, S Sathyamurthy, R Devi, R Jain, JJ Visser. Validation of a commercially available CAD-system for lung nodule detection and characterization using CTscans. Eur Radiol. 2025; 35:1076-1088. Ruitenbeek HC, S Sahil, A Kumar, RK Kushawaha, S Tanamala, S Sathyamurthy, R Agrawal, S Chattoraj, J Paramasamy, D Bos, R Fahimi, EHG Oei, JJ Visser. Cross-validation of an artificial intelligence tool for fracture classification and localization on conventional radiography in Dutch population. Insights Imaging 2025; 16:150.

j.j.visser@erasmusmc.nl

Context

s health care rapidly changes from volume to value-based, there is an urgent need for radiologists to position themselves from a value-based perspective.

Therefore, the Radiology Department at the Erasmus MC has started the value-based imaging program. This means that all activities in the Radiology Department are evaluated in the light of value-based imaging. The value-based imaging program is closely related to the IT department. Most topics that are being covered by this program require expertise in the field of information technology.

225

Research Projects: Objectives & Achievements

- Value driven AI model for detection of incidental pulmonary embolism in CTs (Erik Kemper, Arlette Odink)

Artificial intelligence

- Clinical implementation of Artificial Intelligence solutions for Musculoskeletal Radiology (Huib Ruitenbeek)

- An ethnographic exploration of the digital transformation of healthcare (Gigi Vissers)

Currently, several projects are under investigation in the value-based imaging program. Firstly, as good quality of radiology reports is a core element in the era of valuebased imaging, strategies are developed to investigate the quality of these reports. As manual review to identify sufficient cases for retrospective analysis is often infeasible, a stepwise natural language processing (NLP)method for case identification was developed and applied for the identification of critical finding cases. Quantitative and radiogenomics imaging is becoming more important. It turned out that visual assessment alone was not able to generate all available information. Instead, state-of-the-art computer techniques like machine and deep learning are able to extract quantitative features that are able to play an important role in the diagnostics of musculoskeletal and other diseases. A model was developed and is to be validated that is able to predict the MDM2-amplification in lipoid tumors. Implementation of artificial intelligence algorithms in the radiology workflow is becoming a hot topic. Although several algorithms have been developed of which some are CE-marked / FDA approved and only a few validated in clinical practice, it is important to make the right decisions about implementation of those algorithms. Depending on the type of the algorithm, automatic measurements, case prioritization, pathology detection, a certain amount of research is needed before usage in the clinic can start. Current projects: - AI in radiology and the current legislation framework (Maxime Noordermeer) - Opportunistic screening for osteoporosis in radiological exams using AI (Sybren van Hal) - The impact of AI in health care on professional autonomy (Francesca Guerra) - Understanding the 3D morphology of the calcaneus (Alexander Wakker) - Improving incidental pulmonary nodule detection and follow-up through the implementation of AI-based software (Asabi Leliveld, Arlette Odink, Ties Mulders) - Radiology AI Deployment and Assessment Rubric (RADAR) to bring value-based AI into radiological practice (Bart-Jan Boverhof) - Value impact assessment of AI in radiology (Bina Tariq) - Implementation, cost-effectiveness and carbon footprint of AI when aiming to find lung cancer early (Bo Willems, Arlette Odink, Ties Mulders)

- User-acceptance of AI for radiology in a clinical setting (Jamie Verwey) - Clinical Implementation of AI Technologies for Improved Lung Nodule Detection and Follow-up (Jasika Paramasamy, Arlette Odink, Ties Mulders) - PREDICT_LN: a diagnostic tool with deep-learning model to evaluate thoracic CT scans of colorectal cancer patients for detection and classification of lung nodules (Julie Hamm) - Implementation of artificial intelligence in radiology practice (Laurens Topff)

Data-driven workflow More than 200,000 exams are being performed in our department annually. This requires good logistics and insights in the workflow. In order to get the information about the processes, it is needed to identify relevant data sources. Modern software packages allow for visual representation of process and quality indicators. The research focuses on how to involve people and use dash boarding tools in order to improve efficiency and quality metrics.

Integrated diagnostics Integrated diagnostics will be one of the key steps forward in the coming decade. In this respect, the workup for patients with adrenal incidentaloma was evaluated. Furthermore, the diagnostic work-up for cervix carcinoma, carcinoma of unknown primary, and chest pain was evaluated. Currently, a workflow integration of radiology and pathology findings is investigated in lung cancer patients. Also for liver and brain tumors, further integration of radiology and pathology can increase efficiency and improve diagnosis. In addition, it is to foreseen that image oriented specialties like dermatology can be involved in the integrated diagnostics approach.

Expectations & Directions In the coming years, we aim to further expand our research in the domains of artificial intelligence, datadriven workflow, and integrated diagnostics. An important challenge will be to bring these together and establish reference architectures and pipelines how to structurally evaluate new workflows and technologies.

226

scientific report 2025 | VALUE-BASED IMAGING

Funding Visser, Jacob Evert Stamhuis Convergence: ‘Radiology AI and current legislation’. 2025

Klein, Stefan, Martijn Starmans, Jacob Visser, Kees Verhoef, and Dirk Grunhagen Hanarth Fonds: ‘Automatic grading and phenotyping of soft-tissue tumors through machine learning to guide personalized cancer treatment’. 2021-2025

Visser, Jacob Qure AI: ‘Validation of AI in stroke patients’. 2024-2025

Invited Lectures

Visser, Jacob Qure AI: ‘AI to improve nodule detection on chest X-ray’. 2024-2026 Aerts, Joachim, Jacob Visser AstraZeneca: ‘Pulmonary Incidental Nodules: Improve Detection and Follow-up by integrating Artificial Intelligence (PINPOINT)’. 2024-2028 Visser, Jacob, Martijn Starmans, Ken Redekop, and Frans Vos TKI: ‘Assessing the effect on the clinical workflow and a cost-effectiveness analysis of a comprehensive AI tool for chest CT reading - on the use case of incidental pulmonary embolism’. 2023-2027

Jacob Visser. ‘Legal considerations in Europe regarding radiology AI’. SIIM 2025, Portland, Oregon, USA. May 2025. Jacob Visser. ‘Panel discussion on valuable radiology AI’. Annual Meeting, Crete, Greece. Oct 2025.

Additional Personnel Mart Rentmeester, PhD – ICT

Visser, Jacob Radiology Research Fund: ‘Enabling Value impact assessment for Artificial Intelligence Tools in radiology (eVAIT)’. 2022-2027

Post-docs

Arlette Odink, MD, PhD Pulmonary Incidental Nodules: Improve Detection and Follow-up by integrating Artificial Intelligence in a prospective real-world cohort. To assess the impact of the implementation of CAD combined with dedicated incidental nodule clinics with patient tracking, a prospective observational study will be performed in multiple European medical centres. The rate of stage I and II lung cancer diagnosis with curative intent treatment, health care usage for nodules as well as perform a cost and benefit analysis will be determined. This is de project of a PhD student, and I will be one of the post docs involved.

Carlos Godoy Junior, MD, PhD AIFI Project: Real-World Evaluation and Cost-Effectiveness of AI in Radiology To I am a postdoctoral researcher in the AIFI project, focusing on the real-world evaluation of artificial intelligence in hospital practice. The first AI tool under evaluation within the project is RBfracture, an AI system for fracture detection on radiographs. My work examines its impact on diagnostic performance, clinical workflow, and healthcare resource use in participating hospitals. Using real-world implementation data, I assess the extent to which RBfracture generates measurable clinical gains and whether these gains justify the costs of adoption. This includes estimating the effectiveness of AI implementation in routine care and translating these findings into cost-effectiveness analyses. The overall aim is to provide robust evidence to support decisions on the responsible adoption and reimbursement of AI in radiology.

227

PhD Students Huib Ruitenbeek, MSc

Erik Kemper, MSc

Advisors Jan-Jaap Visser & Edwin Oei

Advisors Jan-Jaap Visser, Martijn Starmans, Ken Redekop & Frans Vos

PhD Obtained: 07-10-2025

Value driven AI model for detection of incidental

Clinical implementation of Artificial Intelligence solutions for Musculoskeletal Radiology

pulmonary embolism in CTs The incidental finding of pulmonary embolisms (IPE) on CTs are quite often missed. The goal is to use a value driven approach to develop, deploy, and evaluate an IPE detection AI model for the radiologic workflow. The deployed model will be part of a comprehensive tool to optimize the clinical workflow for the radiologists, treating physicians and patients.

Artificial Intelligence (AI) solutions are increasingly vital in radiology, aiding or automating tasks to lessen workload. However, clinical evaluation of these algorithms in radiology is scarce. Collaboration between AI vendors and ErasmusMC’s scientific expertise can yield new insights.

Asabi Leliveld, MD, MSc

Jamie Verwey, MSc

Advisors Jan-Jaap Visser & Joachim Aerts

Advisors Sandra Sülz, Jan-Jaap Visser & Maarten IJzerman

Improving incidental pulmonary nodule detection and follow-up through the implementation of AI-based software

User-acceptance of AI for radiology in a clinical setting Understand how factors, values, and actors shape user-acceptance within radiology, and to evaluate the interactions between users and diagnostic AI technology to gain insights into the users’ needs and preferences to further the acceptance and implementation of AI in radiology.

We aim to assess the clinical benefit of implementing computer assisted detection software for nodule detection, combined with a dedicated nodule clinic and patient tracking software. The assessment includes the effect on nodule detection rate, lung cancer stage distribution at diagnosis, healthcare resource use, costbenefit ratio and psychological burden of follow-up.

Jasika Paramasamy, MSc Francesca Guerra, MSc

Advisors Jan-Jaap Visser, Aad van der Lugt & Joachim Aerts

Advisors Yiannis Kyratsis, Maartje Schermer & Jan-Jaap Visser

Clinical Implementation of AI Technologies for Improved Lung Nodule Detection and Follow-up

Algorithmic Technologies shaping Clinical

Artificial Intelligence (AI) is becoming an increasingly important tool in radiology, particularly in lung nodule detection. AI algorithms can assist in identifying nodules on chest radiographs and CT scans, potentially improving detection rates and clinical decision-making. We aim to explore how AI can enhance radiological workflows, improve patient outcomes, and contribute to healthcare efficiency.

Professional Role & Identity The integration of algorithmic technologies, particularly Artificial Intelligence (AI), is transforming the professional role and identity of clinicians. The project aims to understand how AI influences professional autonomy, competence, and moral responsibility in clinical settings, as well as to identify organizational strategies that align AI with clinicians’ role and identity needs.

228

scientific report 2025 | VALUE-BASED IMAGING

Julie Hamm, MSc

Bart-Jan Boverhof, MSc

Advisors Pieter Tanis, Kees Verhoef, Eva Madsen & Jan-Jaap Visser

Advisors Ken Redekop, Maureen Ruttem, Carin Uyl & Jan-Jaap Visser

PREDICT_LN: a diagnostic tool with deep-learning model to evaluate thoracic CT scans of colorectal cancer patients for detection and classification of lung nodules

Radiology AI Deployment and Assessment Rubric (RADAR) to bring value-based AI into radiological practice Radiology artificial intelligence lacks a comprehensive approach to value assessment. To this end, we proposed a comprehensive framework for value assessment of artificial intelligence (AI) in radiology.

Pulmonary nodules in patients with CRC are common and often create diagnostic and therapeutic dilemmas. Improvement of diagnostic tools and classification of pulmonary nodules are necessary. Deep-learning models will be used to evaluate thoracic CT scans.

Bina Tariq, MD, MSc Advisors Meike Vernooij, Jan-Jaap visser & Ken Redekop

Laurens Topff, MD, MSc Advisors Jan-Jaap Visser, Erik Ranschaert & Regina Beets-Tan

Value impact assessment of AI in radiology Evaluation of the added value of artificial intelligence software in clinical practice within the field of radiology and build a value impact assessment framework to facilitate informed decision making on implementation of AI algorithms in clinical practice.

PhD Obtained: 02-04-2025 Implementation of artificial intelligence in radiology practice A data-centric approach to the development and validation of deep learning-based applications for detection and segmentation in chest imaging and neuroradiology.

Bo Willems, MSc

Maxime Noordermeer, LLM

Advisors Joachim Aerts, Jan-Jaap Visser, Kevin ten Haaf & Carlos Godoy

Advisors Jan-Jaap Visser & Evert Stamhuis

Implementation, cost-effectiveness and carbon footprint of AI when aiming to find lung cancer early Within the PINPOINT project, AI tools will aid in the detection and follow-up of Incidental Pulmonary Nodules (IPN). Besides the clinical benefit, it’s important to determine the costeffectiveness and carbon footprint when implemenating AI tools in clinical practice.

ELSA shaping the human-machine interaction with diagnostic AI in radiology Analysis of the ethical, legal and social aspects (ELSA) shaping the human-machine interaction with diagnostic AI in radiology. In this first year, the research focuses on mapping and evaluating the legal and liability framework for diagnostic AI in radiology.

Gigi Vissers, MSc

Sybren van Hal, MD, MSc

Advisors Iris Wallenburg, Jan-Jaap Visser, Petra Porte & Rik Wehrens

Advisors Edwin Oei, Jan-Jaap Visser & Evert van Velsen

An ethnographic exploration of the digital transformation of healthcare: case studies on digital innovations in healthcare and the evolving sociotechnical infrastructure that comprises the social, organisational, technical and ethical aspects of the digitization of care.

Project A.R.G.O.S. (AI-driven Radiology Guidance for Osteoporosis Screening) Validating and exploring the potential impact of AI-based solutions for opportunistic osteoporosis screening.

229

Ryan Muetzel, PhD assistant professor 230

PEDIATRIC POPULATION NEUROIMAGING

Top Publications 2025

JOINT APPOINTMENT IN CHILD AND ADOLESCENT PSYCHIATRY/PSYCHOLOGY

Xu B, H Wang, L Dall’Aglio, M Luo, Y Zhang, RL Muetzel, H Tiemeier. Beyond out-of-sample: robust and generalizable multivariate neuroanatomical patterns of psychiatric problems in youth. Mol Psychiatry 2025; 30:2525-2536.

Ryan Muetzel obtained his BA in Psychology and Biology from the University of Minnesota. After learning neuroimaging at the University of Minnesota’s Center for Magnetic Resonance Research, he joined the Erasmus MC in 2012 for an Msc in Epidemiologyand a PhD in population neuroscience, where he also helped build the Generation R Neuroimaging program. After completing his PhD in 2016, he conducted a postdoctoral fellowship, continued to develop the Generation R Neuroimaging program, and in 2018 became an assistant professor. Ryan now leads the Child and Adolescent Psychiatry department’s pediatric population neuroimaging line and the Generation R Neuroimaging program. Together with his team, the Integrative and Precision Neuroimaging lab, they continue to collect largescale neuroimaging data through the Generation R Study.

Muetzel RL. Enhancing consistency in brain imaging research for population neuroimaging. Nat Protoc. 2025; 20:1099-1100. Kusters MS, L Granés, S Petricola, H Tiemeier, RL Muetzel, M Guxens. Exposure to residential air pollution and the development of functional connectivity of brain networks throughout adolescence. Environ Int. 2025;196:109245.

r.muetzel@erasmusmc.nl

Context

he importance and utility of large-scale pediatric neuroimaging initiatives has become increasingly clear over the last decade. Operating at the intersection of epidemiology child psychiatry, and neuroimaging, the pediatric population neuroimaging program at Erasmus MC has been helping to shine light on the importance of this field, representing one of the first and largest neuroimaging initiatives of its kind in the world. Rich, repeated-measures neuroimaging data collected in tandem with prospective deep-phenotyping data during the prenatal, postnatal, and childhood periods provides unprecedented opportunities to understand neurodevelopment across a multitude of disciplines including neuroscience, psychology, psychiatry, pediatrics, public health, nutrition, and more. My team focuses on understanding typical and atypical neurodevelopment, primarily in the context of emergent psychopathology.

231

Research Projects: Objectives & Achievements My team focuses on 4 key themes. First, typical and atypical neurodevelopment. We use repeated neuroimaging assessments to characterize structural and functional brain changes over time, and examine how this relates various exposures (e.g., early life stress, pollution) as well as outcomes (e.g., psychopathology, substance use initiation). Second, we employ multivariate methods (e.g., machine learning) to identify complex patterns in neuroimaging data, primarily focusing on those which are related to mental health outcomes. These methods allow for the use of high dimensional neuroimaging data in prediction, as well as etiological (explainable) work. Third, we focus on etiological work which examines how various risk and resilience factors relate to the brain and mental health. For example, factors such as physical activity and how it is related to (a)typical brain features. Fourth, my team focuses on developing novel techniques which integrate epidemiological concepts into neuroimaging, or population neuroimaging methods. For example, taking into consideration confounding factors in high dimensional data analysis, how selection bias can be accounted for in large-scale neuroimaging studies, and also applying novel methods such as target trial emulation for research questions which would be challenging to address within an RCT framework.

Understanding Typical and Atypical neurodevelopment Current projects: – Normative Models of the Human Cerebellum (Bing Xu, Carolin Gaiser) – Cerebellar-cortical patterns of neurodevelopment (Carolin Gaiser) Neurodevelopment is a protracted process which begins prenatally and extends into young adulthood. While the last three decades of research have provided an immense amount of information about neurodevelopment from childhood, adolescence and young adulthood, much of the work is based on relatively modest sample sizes and substantial investigation remains a high priority. We have been using the Generation R Study data to better understand typical development in several ways. First, we explored how brain structural connectivity develops over time, and particularly how childhood and pre-adolescent psychiatric problems were related (Dall’Aglio et al.). Using data from both Generation R and the ABCD cohort (Nscans=11,400), we did not demonstrate any robust links between white matter and psychopathology. Given we have identified associations at younger ages, this suggests that earlier periods of development may be more

Figure 1. A.) The Multidomain Task Battery Functional Parcellation of the cerebellum (adapted with permission from J. Diedrichsen). B.) Standardized beta coefficients for age for each functional parcel for males and females separately, and for sex differences. Volume represents the Jacobian determinant, gray/white densities are average tissue class probabilities within a parcel. Data are from 3 time points, over 7,500 scan sessions from 5,000 individuals.

sensitive, or that delayed neurodevelopment can ‘catch up’ with time. Lastly, we developed a normative model of both the anatomical and functional development of the cerebellum (Gaiser et al.). We demonstrated that the cerebellum follows a gradient of development similar to what is observed in the cortex (e.g., sensorimotor areas develop first). Further, we demonstrated that the normative estimates of the cerebellum show robust links with neurodevelopmental disorders (i.e., Autism) at the individual level (i.e., rather than group level).

Risk and Resilience factors of brain and mental health Current Projects: – Air pollution and neurodevelopment (Michelle Kusters – Sex Differences in Autism (Dylan van der Waal, Caro Gaiser) – Maternal Immune Activation and Resilience in Neurodevelopment (Frederieke, Gigase, Milan Zarchev) – Sikkel Cell Disease and Brain Health (Melanie Bruinooge) – Intergeneration Transmission of Mental Health Problems (Boglarka Kovacs)

232

scientific report 2025 | PEDIATRIC POPULATION NEUROIMAGING

Multivariate techniques for neuroimaging Current Project: – Multivariate Prediction of Autism (Carolin Gaiser, Dylan van der Waal, Bing Xu)

Figure 2. Canonical correlation analysis of psychiatric symptoms and brain structure in 10 year-olds. A.) Psychiatric Canonical variate with high Attention Problems loading, B.) Brain-Based Canonical variate, colors depicting brain areas highly correlated to attenion problems.

While several forms of psychopathology have demonstrated considerable heritability, individual-level genomic risk profiles explain only a modest amount of variability in mental health problems. Thus, it remains crucial to explore how various environmental factors related to risk and resilience for mental health. We have been focusing on several key areas, including physical activity, early life adversity, and various prenatal factors such as maternal immune activation. For example, we found that the number of hours of physical activity per week was related to both white matter microstructure cross-sectionally, but also predicted larger neurodevelopmental changes over time in limbic structures known to be both involved in psychopathology and show considerable plasticity. In terms of risk, recent work from our team has shown that early life adversity, both prenatal and during childhood, is related to intracortical myelin. Importantly, contextual risk factors, such as financial difficulties, were a prominent factor in these associations.

233

Neuroimaging in child psychiatric has elucidated several key features of emerging mental illness over the years. However, we still lack robust ‘biomarkers’ which aid in providing families with better information on for example treatment response prediction and longer-term prognosis. Multivariate techniques, such as machine learning and deep learning, may help to uncover more robust, clinically-relevant patterns in neuroimaging data. My group has been applying several “AI” methods in order to better understand whether such features do meaningfully predict clinical information in children and adolescents with, or at risk for, mental illness. Using nearly 13,000 MRI scans from Generation R and ABCD, we found that with both functional MRI data and with structural MRI data, we could detect patterns which suggested a multivariate ‘biotype’ for attention problems in children. However, while functional MRI data from both cohorts produced similar results when run separately, results trained in one cohort did not translate well to the other. Conversely, results from structural MRI data was generalizable across cohorts, suggesting either more robust signal in sMRI or less issues with harmonization of this modality across cohorts. Figure 2 shows canonical variate (psychiatric symptoms with structural brain data) which was robust across cohorts, and model estimates could be transferred from one cohort to the other.

Population neuroimaging methods development Current Projects: – Target Trial Emulation: ADHD Medication and Brain Development (Annet Dijkzeul) – High-dimensional, whole-brain analysis of repeated measures brain data (Serena Defina) In addition to a broad range of descriptive, etiological, and prediction work, my team also develops open-source software to help facilitate population neuroscience research. The clearest example is our implementation of a vertex-wise analysis framework for structural MRI data. We developed the QDECR package (github.com/slamballais/ QDECR) in order to address substantial gaps in current techniques which are particularly relevant for the field of population neuroscience. For example, many current tools have been implemented for much smaller datasets, and do not run properly when the sample size is too large. Further, most packages do not allow for missing data, which can lead to substantial selection bias

instatistical analysis if ignored. Finally, and perhaps most importantly, we ensured the tool was developed in the R statistical software using familiar R code in order to ensure it was user friendly and approachable to neuroimagers without strong backgrounds in statistics and programming. This work has recently led to funding which will allow us to expand the software package to additional statistical methods, and implement new features such as flexibility with data type inputs (e.g., voxels or connectivity matrices instead of vertices)

Muetzel, Ryan, Sander Lamballais, Pauline Jansen et al. FLAG-ERA-JTC: ‘infant2Adult’. 2024-2027 Muetzel, Ryan and Henning Tiemeier EU KA171: ‘Population neuroimaging mobility program - Erasmus MC, Harvard, University of Minnesota’. 2023-2025

Additional Personnel Jessica Wijngaarden - Data manager

Expectations & Directions

Serena Defina - Data Scientist

One of the most important developments of 2025 is the upgrade of the 3T system used by Generation R. This upgrade brings a host of new opportunities for the Pediatric Population Imaging line. The next round of Generation R (focus at 22) began in June of 2025, where we continue our standard protocol, but will also add a specialized functional MRI task designed to realize individualparticipant functional mapping of the brain. The main research priorities for the team remain understanding the neurobiological causes and consequences of mental health problems., With funding awarded for a ZonMw Vidi grant, there will be a particular focus on substance use, and how it is intertwined with neurodevelopment and mental health.. Lastly, our societal outreach via Healthy Start and Convergence with a recent Sprint Award. The project aims to streamline 30 brain printing of individual MRI scans. With a fully-automated pipeline now in place, as well as 3D printing equipment in house, we will begin 3D printing MRI scans and giving them to participants in mid-2026 as a “thank you” for their participation in the MRI study.

Funding

Jana Hermans - PhD student, Department of Child and Adolescent Psychiatry/Psychology Dogukan Koc - PhD student, Department of Child and Adolescent Psychiatry/Psychology Tessa Mulder - PhD student, Department of Internal Medicine Laura Monteiro - PhD student, Department of Child and Adolescent Psychiatry/Psychology Merel de Vries - PhD student, Department of Adult Psychiatry Shuaiqun Pan - PhD student, Computer Science, Leiden University Amina Burgess - PhD student, Spinoza Center for Neuroimaging Sharon Ololo - Researcher, Department of Adult Psychiatry

Muetzel, Ryan NIH / NIMH - IMPACT Study: ‘Understanding Maternal Immune Activation and child neurodevelopment’. 2021-2026

Jin Ouyang- MSc Student, TU Eindhoven Remmy Duijsens - MSc Student, TU Delft

Muetzel, Ryan, Maarten Frens, Alexandra Badura, Rick van der Vliet, Jeremy Labrecque, and Monique Bernsen Erasmus MC Synergy: ‘Understanding Cortico-cerebellar maturation trajectories’. 2023-2026

Hannah van Heusden - MSc Student, Medicine, Erasmus MC

Muetzel, Ryan, Jeremey Labreqcue, and Henning Tiemeier Sophia Foundations: ‘ADHD, medication use and brain development: Using target trial emulation to infer the effects of stimulant medication’. 2021-2025 Badura, Aleksandra, Ryan Muetzel et al. NWO NWA ORC: ‘Why are there more men than women with autism? Sex differences in Autism: Genes, Brain, and Healthcare’. 2024-2029 Diedrichsen, Jorn, Andre Marquand, and Ryan Muetzel Raynor Cerebellum Project: ‘Normative growth models of the human Cerebellum’. 2024-2027

234

scientific report 2025 | PEDIATRIC POPULATION NEUROIMAGING

Post-docs

Milan Zarchev, PhD Early Life Adversity, Maternal Immune Activation, and Neurodevelopmental Outcomes Early-life adversity has been consistently shown to impact neurodevelopment, including pubertal timing, behavioural and emotional problems such as ADHD, and also brain morphology and function. However, inconsistencies exist in the literature, suggesting mostly very severe forms of adversity affect neurodevelopment. This project aims to a.) utilize novel causal inference methods to understand how ELA links with neurodevelopment, and b.) particularly test whether maternal immune activation profiles are linked to adverse neurodevelopment.

Frederieke Gigase, PhD Maternal Immune Activation, and Neurodevelopmental Outcomes Maternal immune activation has been shown to be considerably distinct from both pre- and post-pregnancy immune profiles. Decades of work has pointed to adverse neurodevelopmental outcomes linked to maternal immune profiles. This has been particularly linked to maternal infections during pregnancy. However, various factors not linked to infections also have been shown to impact immune profiles, for example obesity, smoking, and early life stress. This project aims to examine in detail whether there is a link between the maternal immune profile and the risk for offspring to develop autism. This project will also ex-plore how the maternal immune profile is linked to other neurodevelopmental outcomes, such as brain structure and function.

PhD Students Bing Xu, MSc

Michelle Kusters, MD, MSc

Advisors Ryan Muetzel & Henning Tiemeier

Advisors Ryan Muetzel, Henning Tiemeier & Monica Guxens

Michelle works on a collaborative PhD project between ISGlobal in Barcelona and the Erasmus MC, which focusses on the influence of environmental pollutants on neurodevelopment in children. Her project has focused specifically on understanding the potential consequences of air pollution on behavior, cognition, and brain development.

Bing is a PhD student with a background in developmental and computational psychology. She is interested in applying advanced statistical techniques in neuroimaging to address the heterogeneity in child psychopathology. She has been working on identifying ‘biotypes’ of children’s ADHD via brain connectivity and morphology using machine learning.

235

Caro Gaiser, MSc

Boglarka Kovacs, MSc

Advisors Ryan Muetzel, Rick van der Vliet & Maarten Frens

Advisors Neeltje van Haren, Alex Neumann, Lisanne van Houtum & Ryan Muetzel

Caro is a PhD student investigating the cerebellum, motor control and motor learning at the Department of Neuroscience and in the Generation R study. Her research focuses on describing typical cerebellar development and finding factors that can explain individual differences in motor learning using behavioral, electrophysiological, neurostimulation, and neuroimaging methods.

The FAMILY Project This project aims to characterize if variations in brain metrics constitute plausible pathways from intergenerational risk of mental illness from parent to offspring. Further, Boglarka will elucidate whether identified neuroimaging biomarkers improve the prediction of mental health outcomes in persons at familial high-risk for mental illness.

Eva Borkhuis, MSc Advisors Ryan Muetzel, Danielle Remmerswaal, Ruth van der Hallen & Pauline Jansen

Melanie Bruinooge, MSc Advisors Marjon Cnossen, Ryan Muetzel & Sabine Mous

Eva is a PhD student working within the convergence initiative, namely within Healthy Start. Her project is situated within the mental well-being ambition of healthy start, and focuses on understanding risk and resilience factors of mental wellbeing using a novel, youth-centered approach.

The BRICK Study Sickle cell disease (SCD) is a recessive red blood cell blood disorder. Individuals with SCD have an increased risk of both overt cerebral infarctions and silent infarctions. This project maps whether brain tissue damage precedes, parallels, or lags behind non-brain tissue damage in youth with SCD.

Annet Dijkzeul, MSc Advisors Ryan Muetzel, Jeremy Labrecque & Henning Tiemeier

Dylan van der Waal, MSc

Annet is a PhD student with a background in Neuropsychology and Neurosciences. Her research focuses on combining neuroimaging and epidemiological techniques to better understand neurodevelopment and ADHD. Specifically, she will utilize target trial emulation to make causal inferences on the impact of ADHD medication on brain development.

Advisors Ryan Muetzel, Pauline Jansen & Alexandra Badura

The SCANNER project The aim within SCANNER is to identify brain features that co-segregate with sex in autism. Large, population-based data (Generation R) will allow us investigate the correlation between structural and functional brain changes over development, autistic traits and sensorimotor difficulties, particularly in the context of sex differences.

Anna Suleri, MSc Advisors Charlotte Cecil, Tonya White & Veerle Bergink Anna is a PhD student working on an NIH-funded R01 grant to better understand how prenatal exposure to maternal infection and immune activation are related to neurodevelopment. She is focusing on self-reported measures of infection status during pregnancy and a unique cytokine assay of samples collected during early pregnancy.

236

scientific report 2025 | PEDIATRIC POPULATION NEUROIMAGING

237

238

INPUT & OUTPUT

239

CONFERENCE CONTRIBUTIONS 2025 (selection)

This section provides a selection of the 2025 conference contributions from the Erasmus MC Department of Radiology & Nuclear Medicine. Given the large number of conferences to which we contribute, this overview is necessarily incomplete, but it gives some idea of the magnitude of our conference participation. Conference

City

Country/States

Number

ARVO

Salt Lake City

USA

Cancer Retreat

Rotterdam

Netherlands

s Hertogenbosch

The Netherlands

Dutch Federated Learning Symposium

DDRC

Amsterdam

The Netherlands

EANM

Barcelona

Spain

EANO

Praque

Czech Republic

EASL

Amsterdam

The Netherlands

EMIM

Bilbao

Spain

ESMRMB

Marseille

France

ESGAR

Amsterdam

The Netherlands

ESNR

Istanbul

Turkiye

ESTRO

Vienna

Austria

European Congress of Radiology

Vienna

Austria

European Stroke Organisation Conference

Helsinki

Finland

ISCFS

Shanghai

China

ISBI

Houston

Texas

ISGDR

Oslo

Norway

ISMRM

Honolulu

Hawaii

ISMRM BEnelux chapter

Antwerpen

Belgium

ISMRM BeNelux Chapter

Hilversum

The Netherlands

ISMRM Perfusion workshop

Pamplona

Spain

IWOAI

Cambridge

England (UK)

MICCAI

Daejon

South Korea

NKRV

Rotterdam

Netherlands

NOGC

Maastricht

The Netherlands

OHBM

Brisbane

Australia

ORS

Phoenix

Polish Society of Ophthalmology (PTO) Meeting

Krakow

Poland

PTO

Braga

Portugal

WCC

Washington, DC

USA

World Stroke Conference

Barcelona

Spain

WROP

Florence

Italy

Total

240

Continent

Countries

Australia

Country/States

Contributions

1 Australia

Europe

27 Austria Belgium Czech Republic England Finland France Italy Norway Poland Portugal Spain The Netherlands

Asia

4 China South Korea Turkiye

USA

5 Arizona Hawaii Texas Washington D.C. Utah

241

PUBLICATIONS 2025 Publicationlist To see compete publicationlist of all our staffmembers and external publication partners, please scan QR-code.

PhD Dissertations 1.

Stijntje Dijk. 2025, February 11. Decision-Making in Medical Research, Practice and Education. Promotor/co-promotor: Prof. dr. M.G.M. Hunink, Prof. dr. J.B. Wong, Dr. J.A. Labrecque.

Karlijn de Joode. 2025, March 4. Optimizing the Individualized Treatment Strategy for Immune Checkpoint Inhibitors in Patients with Cancer. Promotor/copromotor: Prof. dr. A.J.H. Mathijssen, Dr. A.A.M. van der Veldt.

Laurens Topff. 2025, April 2. Bridging the AI Gap in Radiology: From Data to Clinical Impact. Promotor/co-promotor: Prof. dr. R.G.H. Beets-Tan, Prof. dr. E.R. Ranschaert, Dr. J.J. Visser.

Jacqueline Claus. 2025, April 15. Stroke and Neuroimaging as Indicators for Dementia Risk Towards personalized prevention and targeted therapies. Promotor/ co-promotor: Prof. dr. M.A. Ikram, Prof. dr. M.W. Vernooij, Dr. F.J. Wolters.

Xianjing Liu. 2025, May 14. 3D Facial Shape Analysis Unraveling environmental genetic, and health connections through AI-driven methodological advancements. Prof. dr. E.B. Wolvius, Dr. G.V. Roshchupkin.

Nikki Boodt. 2025, May 27. Clinical Relevance of Thrombus Characterization in Acute Ischemic Stroke Histological, mechanical and radiological analyses. Promotor/copromotor: Prof. dr. A. van der Lugt, Prof. dr. D.W.J. Dippel, Prof. dr. H.F. Lingsma.

242

Sophie Derks. 2025, June 24. Gain in the brain: Improving diagnosis and treatment for patients with brain metastases. Promotor/copromotor: Prof. dr. M. Smits, Prof. dr. M.J. van den Bent, Dr. A.A.M. van der Veldt.

Aikaterini Tziotziou. 2025, August 26. Coronary and Carotid Atherosclerosis: Morphometrics and Biomechanics. Prof. dr. ir. A.F.W. van der Steen, Prof. dr. A. van der Lugt, Dr. A.C. Akyildiz, Dr. D. Bos.

Róisín McMorrow. 2025, August 27. Harnessing the Light: Therapy and optical imaging for pancreatic cancer. Promotor/co-promotor: Prof. dr. C.W.G.M. Lowik, Dr. L. Mezzanotte.

10. Huib Ruitenbeek. 2025, October 7. Validation and impact of AI tools for fracture detection in radiology. Promotor/co-promotor: Prof. dr. E.H.G. Oei, Dr. J.J. Visser.

11. Merel de Vries. 2025, October 29. Plasticity of the Maternal Brain: A lifespan perspective. Promotor/ co-promotor: Prof. dr. S.A. Kushner, Prof. dr. M.W. Vernooij, Dr. E.M.P. Poels.

12. José Castillo Tovar. 2025, November 11. Computer aided diagnosis for Prostate Cancer classification on multiparametric MRI: Generalization and validation of radiomics and Deep learning approaches. Promotor/copromotor: Prof. Dr. W.J. Niessen, Dr. J.F. Veenland, Dr. I.G. Schoots.

13. Eline Hooijman. 2025, November 26. The Clinical Translation of Targeted Radionuclide Therapy with Lu-177 and Ac-225: From Radionuclide to Standard of Care. Promotor/co-promotor: Prof. dr. F.A. Verburg, Prof. dr. P.H.M. van der Kuy, Dr. S.L.W. Koolen, Dr. R.H. de Blois.

14. Abdullah Thabit. 2025, December 9. Unbounded Surgical Reality Development and assessment of augmented reality navigation for image-guided surgery. Promotor/ co-promotor: Dr. T. van Walsum, Prof. dr. E.B. Wolvius.

15. Angelina Pieters. 2025, December 18. Image Analysis and Treatment of Bronchiectasis. Promotor/ co-promotor: Prof. Dr. H.A.W.M. Tiddens, Dr. M.M. van der Eerden, Dr. P. Ciet.

243

INDEX

A Alafandi ........................................................ 157 Alim ................................................................ 157 Aliukonyte ................................................... 201 Andrade de Jesus .......................................... 96 Arnaud ........................................................... 136 Asuzu ............................................................ 157 Audekerke ....................................................... 75 Audhoe .......................................................... 143 B Bastiaanse ....................................................... 70 Becx ............................................................... 111 Benmahdjoub ................................................ 56 Bie ...................................................................223 Birhanu ............................................................ 94 Blois ............................................................... 135 Bocharewicz ................................................. 195 Boeren .......................................................... 188 Bokhout ........................................................ 117 Bontempi ..................................................... 223 Booij ................................................................. 45 Boodt ............................................................ 150 Borkhuis ....................................................... 236 Bos ................................................................ 218 Bortsov ............................................................ 93 Boverhof ....................................................... 229 Box ................................................................. 217 Braak .............................................................. 179 Brabander ..................................................... 124 Bracke ............................................................ 210 Bron .................................................................. 76 Bruijne ............................................................. 46 Bruinooge ..................................................... 236 Budde ............................................................ 164 C Caudri............................................................. 196 Campeiro....................................................... 116 Chapeau......................................................... 136 Chen.................................................................. 63 Cheung........................................................... 143 Ciet.................................................................. 196 Claus............................................................... 217 Coerts............................................................. 123

Cornelissen.................................................. 194 Cox................................................................. 128 Cretu ............................................................. 201 D Dalebout ...................................................... 171 Dalm ............................................................. 112 Damiana ....................................................... 117 Deng ............................................................. 189 Derks ............................................................ 128 Dias ................................................................. 83 Dijk ................................................................ 171 Dijkzeul ........................................................ 236 Dille .............................................................. 157 Dremmen ..................................................... 211 Droogers ...................................................... 143 Dur ................................................................ 189 Dwarskasing ............................................... 177 E Engbers ........................................................ 111 F Faure ............................................................. 171 G Gahrmann .................................................... 156 Gaiser ........................................................... 236 Garrido Hernandez .................................... 109 Gigase .......................................................... 235 Godoy Junior .............................................. 227 Graaf ............................................................. 150 Guerra .......................................................... 228 H Haaften ........................................................... 83 Hal ................................................................. 229 Hamm ........................................................... 229 Hanff ............................................................. 180 Hartmann ....................................................... 70 Heide ............................................................ 116 Heijden ........................................................ 187 Heredia ........................................................ 110 Hernandez Tamames ................................... 34 Hermann......................................................... 70

244

Heuvel .......................................................... 188 Hilten ............................................................... 63 Hirsh .............................................................. 164 Hirvasniemi .................................................... 84 Hooijman ...................................................... 136 Houwelingen ............................................... 211 I Ijtsma ............................................................. 123 Ivanov ............................................................ 123 J Jong .............................................................. 143 Jonge ............................................................... 83 K Kaaij ............................................................... 188 Kamphuis ........................................................ 87 Kang ................................................................. 83 Kayal ................................................................ 49 Kemper, E ...................................................... 228 Kemper, L ...................................................... 162 Kemmeren .................................................... 189 Khan ............................................................... 223 Kilinc .............................................................. 211 Klein ................................................................ 64 Kleijn ............................................................. 111 Klomp ............................................................ 117 Konijnenberg ............................................... 121 Kotek ................................................................ 38 Kovacks ......................................................... 236 Kruijff ............................................................ 137 Kusters .......................................................... 235 L Laan .................................................................. 45 Lavrova .......................................................... 162 Lee ................................................................. 128 Leening ......................................................... 222 Leeuwen ....................................................... 157 Leliveld ......................................................... 228 Li, B ................................................................... 82 Li, L ................................................................. 137 Li, X ................................................................ 223 Li, Z ................................................................... 63

Lima ............................................................... 111 Ling ................................................................ 128 Liu, G ................................................................ 75 Liu, X ................................................................ 63 Loddin ........................................................... 223 Lugt .................................................. 4, 147, 190

Q Qian ................................................................. 94 R Raut .............................................................. 201 Rojas ................................................................ 70 Roshchupkin .................................................. 58

M Ma ................................................................... 136 Maijoor .......................................................... 189 Marzetti ........................................................... 94 Menart ........................................................... 217 Meyer Viol ....................................................... 45 Mo ..................................................................... 82 Mollica ........................................................... 201 Mostert .......................................................... 188 Mouheb ........................................................... 83 Muetzel ......................................................... 230 Mulders ......................................................... 122 N Neitzel ........................................................... 209 Neri ................................................................... 40 Neves ............................................................. 100 Nguyen .......................................................... 211 Niemandsverdriet ......................................... 70 Nijenhuis ......................................................... 57 Nonnekens ....................................................104 Noordermeer ............................................... 229 Ntihabose ..................................................... 137 O Odink ............................................................. 227 Oei .................................................................. 180 Orsi ................................................................ 189 Overdevest ................................................... 143 Oviedo Acosta ................................................ 94 Ozdamar ....................................................... 128 P Paassen ......................................................... 189 Paraiso ........................................................... 137 Paramasamy ................................................. 228 Perrin ............................................................. 109 Pieterman ..................................................... 194 Pieters ........................................................... 200 Pomp ............................................................. 217 Poot .................................................................. 72 Prathaban ........................................................ 94 Pruis ............................................................... 142

Rosbergen ................................................... 211 Roumen .......................................................... 57 Ruitenbeek .................................................. 228 S Sabatella ...................................................... 110 Sabidussi ........................................................ 75 Samadifardheris ........................................... 75 Sambou ........................................................ 217 Sanchez Brea ................................................. 96 Sanden ......................................................... 123 Schallig ........................................................ 186 Schoon ......................................................... 150 Schoots ........................................................ 172 Seimbille ..................................................... 130 Seyrek .......................................................... 179 Sharma ......................................................... 171 Sgambelluri ................................................... 40 Silva Pereira Fialho da Peidade ................ 40 Sluijs ............................................................... 56 Smits ............................................................. 152 Spaanderman ................................................ 71 Starmans ........................................................ 88 Straten ............................................................ 42 Streiber ........................................................ 223 Suleri ............................................................ 236

Veenstra ....................................................... 179 Vegt ................................................................ 121 Veldhuijzen van Zanten ............................ 138 Veldt ............................................................... 124 Verburg .......................................................... 118 Vernooij ......................................................... 204 Verwey ........................................................... 228 Vinke .............................................................. 210 Vissers ........................................................... 229 Visser ............................................................. 224 Vries ............................................................... 210 Vreede ........................................................... 188 Vrooman .......................................................... 81 Venugopal ....................................................... 40 W Waal ............................................................... 236 Wakker ............................................................. 57 Walsum ............................................................ 50 Wang ................................................................ 75 Wan .................................................................. 94 Warnert ......................................................... 158 Weijden ........................................................... 39 Werff ................................................................ 75 Weustink ....................................................... 170 Wijngaarden ................................................... 83 Willemssen .................................................. 179 Willems ......................................................... 229 Wolff ................................................................ 57 Wolf, M .......................................................... 195 Wolf, T ........................................................... 111 Wolfert .......................................................... 123 Wolters .......................................................... 212 X Xu ................................................................... 235

T Tamborino .................................................. 108 Tang, J ............................................................. 57 Tang, P .......................................................... 162 Tariq .............................................................. 229 Thabit .............................................................. 56 Thomeer ...................................................... 178 Tohidineshad ................................................. 93 Topff .............................................................. 229 Tziotzio ......................................................... 222 V Veer ............................................................... 201 Veerdonk ..................................................... 156 Veenland ........................................................ 69

245

Y Yu ...................................................................... 63 Z Zadelhoff ...................................................... 189 Zarchev ......................................................... 235 Zerguit ........................................................... 123 Zang ................................................................. 40 Zijta .................................................................. 71 Zoetelief ....................................................... 128

246

SOCIAL ACTIVITIES

247

248