International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
AI-Assisted Diagnostic Framework for Early Brain Hemorrhage Recognition from CT Imaging Chetankumar Kalaskar1,Imran Ali2 ,Abdul Muqeeth3,Akhlaq Ahmed4 1Assitant Professor,Computer Science & Engineering,PDA College,Kalaburagi,Karnataka,India 2-4 Students, Computer Science & Engineering ,PDA College,Kalaburagi, Karnataka,India
------------------------------------------------------------------------***-------------------------------------------------------------------------
Abstract- Brain hemorrhage, a critical medical emergency
transfer learning techniques, enabling the system to learn visual patterns corresponding to different hemorrhage types. Unlike traditional machine learning models that require handcrafted features, deep learning extracts hierarchical visual representations directly from data, improving sensitivity to subtle abnormalities. A web-based interface built using Flask allows seamless interaction, enabling users to upload CT images and receive classification results, while a backend database manages prediction history for review and research continuity. The system is designed to be scalable, user-friendly, and suitable for integration into real-time clinical workflows, especially in regions with limited specialist availability. By combining automation, accessibility, and accuracy, the framework aims to enhance diagnostic efficiency, assist medical professionals under high workload conditions, and contribute to faster decision-making in emergency care.
caused by bleeding within the brain tissue, demands rapid and accurate diagnosis to prevent severe neurological damage or fatal outcomes. Manual interpretation of CT scans is timeconsuming and highly dependent on radiologist expertise, which creates the need for automated, assistive diagnostic systems. This study presents an AI-assisted framework designed to analyze brain CT images and identify hemorrhagic conditions at an early stage. The approach incorporates preprocessing, segmentation, and a transfer-learning-based deep neural model trained to classify CT scans as hemorrhagic or non-hemorrhagic with high reliability. A Flask-based web interface enables real-time image upload, prediction, and result visualization, supporting practical clinical integration. The system also stores processed data for further audit and research extensibility. Experimental evaluation demonstrates strong performance across accuracy, sensitivity, and specificity metrics, indicating potential for deployment in screening workflows, especially in regions lacking immediate radiology access. This framework contributes to faster decision support, reduced diagnostic load, and improved patient triage in emergency care.
2. RELATED WORKS Article [1] 'Detection of Intracranial Hemorrhage by Artificial Intelligence' by D. Veselov et al. in 2024: This paper addresses the critical task of intracranial hemorrhage detection and classification using deep learning approaches applied to the RSNA 2019 brain computed tomography dataset. The authors developed an artificial intelligence system capable of analyzing CT scans to identify hemorrhagic patterns with high precision. The methodology incorporates advanced convolutional neural network architectures trained on over 25,000 annotated CT images, enabling the model to learn hierarchical visual features associated with different hemorrhage subtypes. The study demonstrates that automated detection systems can significantly reduce diagnostic time while maintaining accuracy comparable to expert radiologists.
Keywords: Brain Hemorrhage, CT Imaging, Deep Learning, Medical Image Classification, AI-Assisted Diagnosis, Transfer Learning, Automated Screening, Clinical Decision Support
1.INTRODUCTION Brain hemorrhage is a life-threatening neurological condition that occurs when blood vessels rupture and leak within the brain, leading to increased intracranial pressure, loss of oxygen supply, and rapid tissue damage. Early identification plays a vital role in improving patient survival, yet diagnosis often depends on the availability of expert radiologists and timely analysis of CT scans. In emergency settings, delays in assessment can significantly worsen clinical outcomes. This challenge has accelerated interest in computer-aided systems that can support medical professionals by automating hemorrhage detection, reducing interpretation time, and minimizing human error.
Article[2] 'A Deep Learning Algorithm for Automatic Detection and Classification of Intracranial Hemorrhage' by X. Wang et al. in 2021: This influential study presents a comprehensive deep learning algorithm developed using the extensive 2019-RSNA Brain CT Hemorrhage Challenge dataset containing over 25,000 CT scans with diverse hemorrhagic presentations. The authors employed transfer learning techniques combined with data augmentation strategies to address class imbalance issues commonly encountered in medical imaging datasets.The proposed algorithm demonstrates exceptional performance in both binary classification
The project focuses on developing an AI-assisted diagnostic framework that analyses brain CT images and distinguishes between hemorrhagic and non-hemorrhagic cases. The methodology integrates image pre-processing, region segmentation, and deep learning-based classification using
© 2025, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 893