Enhancing Pediatric Brain Tumor Segmentation through Transfer Learning from Adult Data by IRJET Journal

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 11 Issue: 09 | Sep 2024

p-ISSN: 2395-0072

www.irjet.net

Enhancing Pediatric Brain Tumor Segmentation through Transfer Learning from Adult Data A.Srujan Reddy

A. Sai Veera Manikanth

Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India.

R .Venkata Adi Rama Pardhu Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India.

A.Siva Naga Sai Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India.

-----------------------------------------------------------------------***-------------------------------------------------------------------------Abstract---In order to improve accuracy and reliability, this study presents a unique method for segmenting juvenile brain tumors by using transfer learning from adult brain tumor datasets. Our strategy outperforms conventional approaches in segmentation by fine-tuning pre-trained deep learning models on a pediatric-specific dataset. Better patient outcomes, diagnosis, and treatment planning result from this. Our method may completely change how juvenile brain cancers are identified and treated, leading to better results for affected patients.

experimental findings show a considerable increase in segmentation performance compared to traditional approaches. Better therapeutic results are the end result of this method's ability to increase segmentation accuracy while also easing the process of early diagnosis and individualizing treatment plans for pediatric patients. Transfer learning for pediatric brain cancers is a revolutionary approach that improves the accuracy of segmentation for uncommon children brain tumors by borrowing information from other types of brain tumors, namely adult brain tumors. MRI, deep learning, and transfer learning are used in Pediatric Brain Tumor Segmentation, a novel method that addresses the difficulties in accurately segmenting tumors in children's brains.

Keywords-- Pediatric brain tumors, transfer learning, deep learning, MRI segmentation, adult brain tumor data, medical imaging, fine-tuning, personalized treatment, early diagnosis, neural networks.

This approach improves accuracy and efficiency by finetuning pre-trained deep learning models on huge adult brain tumor datasets to adapt to the particular features of pediatric brain tumors. Brain tumors can be accurately segmented from MRI images by applying advanced filtering and thresholding techniques. Better diagnostic and treatment planning, more individualized care, and ultimately better patient outcomes are the results of this strategy. The technical method is gathering and annotating a dataset that is particular to children, choosing and optimizing a pre-trained model, training and assessing the model, and applying the model to identify brain tumors in fresh MRI pictures. Future initiatives involve investigating novel deep learning architectures to further increase segmentation efficiency and accuracy, integrating with clinical workflows, and conducting multi- center research to improve the generalizability of the model. Pediatric Brain Tumor Segmentation has the potential to transform the diagnosis and treatment of pediatric brain tumors,

Introduction

This research introduces a new method that uses transfer learning from data on adult brain tumors to enhance the segmentation of tumors in children's brains. Traditional techniques of precise segmentation are challenged by the rarity and diversity of pediatric brain tumors. We enhance the accuracy and reliability of these models by adapting deep learning models that were pre-trained on large adult brain tumor datasets to the pediatric scenario. Our approach includes adjusting the pre-trained models using a pediatric- specific dataset, which maintains important traits while taking into account the specifics of pediatric cancers. Highlighting the potential of transfer learning to solve the limits of pediatric brain tumor analysis, the

Impact Factor value: 8.315

ISO 9001:2008 Certified Journal

Page 781