Development of a CAD system for stroke diagnosis using machine learning on DWI-MRI images
Stroke remains one of the leading causes of disability and mortality worldwide, necessitating timely and accurate diagnosis to improve treatment outcomes. This study presents a computer-aided diagnosis (CAD) system designed to detect and classify stroke lesions in magnetic resonance imaging (MRI),...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Accent Social and Welfare Society
2025
|
| Online Access: | http://eprints.utem.edu.my/id/eprint/29556/2/00332220920251135222149.pdf http://eprints.utem.edu.my/id/eprint/29556/ https://accentsjournals.org/PaperDirectory/Journal/IJATEE/2025/8/5.pdf http://dx.doi.org/10.19101/IJATEE.2024.111102264 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Stroke remains one of the leading causes of disability and mortality worldwide, necessitating timely and accurate
diagnosis to improve treatment outcomes. This study presents a computer-aided diagnosis (CAD) system designed to detect and classify stroke lesions in magnetic resonance imaging (MRI), specifically utilizing diffusion-weighted imaging (DWI) sequences. A hybrid segmentation technique, fuzzy c-means with active contour (FCMAC), is proposed to enhance lesion localization accuracy. For classification, the system evaluates traditional machine learning algorithms like support vector machine (SVM) and k-nearest neighbor (KNN), alongside deep learning models such as convolutional neural
network (CNN) and bilayered neural network (BNN). The entire diagnostic pipeline is integrated into a MATLAB-based
graphical user interface (GUI), facilitating real-time analysis and ease of use in clinical settings. Experimental results
show that the proposed FCMAC method achieves a dice coefficient (DC) of 0.654, outperforming conventional
segmentation techniques. Among the classifiers, KNN offered the best balance between prediction accuracy and
computational efficiency. The final system, termed SmartStroke-Pro, enables early detection and classification of stroke, providing a reliable and practical tool to assist healthcare professionals, particularly in resource-limited environments. This framework has the potential to reduce diagnostic delays and support improved clinical decision-making in acute stroke care. |
|---|