View in EDS

Deep Learning-Based Classification of Stroke MRI Scans using Convolutional Neural Networks (CNNs)

Saved in:
Bibliographic Details
Title: Deep Learning-Based Classification of Stroke MRI Scans using Convolutional Neural Networks (CNNs)
Authors: Sambhav Gupta
Source: International Journal of Sciences and Innovation Engineering; Vol. 2 No. 6 (2025): IJSCI VOLUME-02 ISSUE-06 JUNE 2025; 831-839 ; 3049-0251 ; 10.70849/
Publisher Information: International Journal of Sciences and Innovation Engineering
Publication Year: 2025
Subject Terms: Stroke Classification, Deep Learning, CT Scans, Convolutional Neural Networks, Image Augmentation, Medical Imaging
Description: Stroke remains one of the leading causes of death and long-term disability globally [1], [2]. Rapid and accurate classification of stroke type ischemic or hemorrhagic is essential for timely medical intervention, as treatment approaches vary significantly between types [3]. While non-contrast computed tomography (CT) is the standard diagnostic imaging tool due to its speed and availability, interpretation relies on expert radiologists, which may not be feasible in all clinical settings, particularly in rural or resource-limited areas [4], [5].In this study, we present a deep learning-based model using Convolutional Neural Networks (CNNs) for the automated classification of brain CT images into three categories: Ischemic Stroke, Hemorrhagic Stroke, and Normal Brain. A curated dataset of 1,865 CT scan images was preprocessed using resizing, normalization, and data augmentation techniques to improve generalization. The CNN was trained from scratch using the Keras framework and achieved a validation accuracy of 66%. Performance was further evaluated using precision, recall, F1-score, and confusion matrix analysis. Although the model showed strong performance for normal brain classification, it exhibited lower recall for hemorrhagic stroke cases, indicating the need for architectural optimization or transfer learning techniques in future iterations. Nevertheless, this study highlights the feasibility of deploying AI-powered diagnostic tools to assist radiologists, improve diagnostic speed, and extend accessibility to stroke care in underserved regions [6]–[13].
Document Type: article in journal/newspaper
File Description: application/pdf
Language: English
Relation: https://ijsci.com/index.php/home/article/view/302/264; https://ijsci.com/index.php/home/article/view/302
DOI: 10.70849/IJSCI
Availability: https://ijsci.com/index.php/home/article/view/302
https://doi.org/10.70849/IJSCI
Rights: https://creativecommons.org/licenses/by-nc/4.0
Accession Number: edsbas.BC190CA8
Database: BASE
Description
Abstract:Stroke remains one of the leading causes of death and long-term disability globally [1], [2]. Rapid and accurate classification of stroke type ischemic or hemorrhagic is essential for timely medical intervention, as treatment approaches vary significantly between types [3]. While non-contrast computed tomography (CT) is the standard diagnostic imaging tool due to its speed and availability, interpretation relies on expert radiologists, which may not be feasible in all clinical settings, particularly in rural or resource-limited areas [4], [5].In this study, we present a deep learning-based model using Convolutional Neural Networks (CNNs) for the automated classification of brain CT images into three categories: Ischemic Stroke, Hemorrhagic Stroke, and Normal Brain. A curated dataset of 1,865 CT scan images was preprocessed using resizing, normalization, and data augmentation techniques to improve generalization. The CNN was trained from scratch using the Keras framework and achieved a validation accuracy of 66%. Performance was further evaluated using precision, recall, F1-score, and confusion matrix analysis. Although the model showed strong performance for normal brain classification, it exhibited lower recall for hemorrhagic stroke cases, indicating the need for architectural optimization or transfer learning techniques in future iterations. Nevertheless, this study highlights the feasibility of deploying AI-powered diagnostic tools to assist radiologists, improve diagnostic speed, and extend accessibility to stroke care in underserved regions [6]–[13].
DOI:10.70849/IJSCI