Related Experiment Video
Updated: Jun 27, 2025
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
48.0K
Automated Ischemic Stroke Classification from MRI Scans: Using a Vision Transformer Approach
Wafae Abbaoui1,2, Sara Retal3, Soumia Ziti1
1Intelligent Processing & Security of Systems (IPSS) Research Team, Faculty of Sciences, Mohammed V University in Rabat, Rabat 10000, Morocco.
Journal of Clinical Medicine
|April 27, 2024
Summary
The Vision Transformer (ViT-b16) model achieved 97.59% accuracy in classifying ischemic stroke from MRI scans, significantly outperforming the VGG-16 model. This highlights advanced AI for improved stroke diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Evaluating deep learning models for medical image analysis.
- Focusing on ischemic stroke classification using MRI scans.
- Utilizing Moroccan patient data for model training and validation.
Purpose of the Study:
- To assess the performance of the Vision Transformer (ViT-b16) model.
- To compare ViT-b16 against the Visual Geometry Group 16 (VGG-16) model.
- To determine the efficacy of ViT-b16 in classifying ischemic stroke from MRI scans.
Main Methods:
- Dataset compilation of 342 MRI scans (Normal vs. Stroke).
- Image preprocessing using TensorFlow's tf.data API.
- Training and evaluation of the ViT-b16 model.
Main Results:
- ViT-b16 achieved an accuracy of 97.59%.
- ViT-b16 significantly outperformed the VGG-16 model (90% accuracy).
- Demonstrated superior classification capabilities for ischemic stroke.
Conclusions:
- ViT-b16 shows high potential for accurate ischemic stroke diagnosis.
- Advanced deep learning models can enhance medical image analysis.
- Supports further research into AI for improved clinical healthcare outcomes.
