Optimizing stroke detection with genetic algorithm-based feature selection in deep learning models

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Brain stroke is a major cause of death and disability globally.
• Accurate and efficient diagnostic tools are crucial for timely intervention.
• Current diagnostic methods can be time-consuming and require specialized expertise.

Purpose of the Study:

• To enhance stroke detection accuracy using deep learning models.
• To integrate Genetic Algorithms (GA) for optimized feature selection in neuroimaging.
• To evaluate the performance of InceptionV3, VGG19, and MobileNetV2 for stroke classification.

Main Methods:

• Neuroimaging data classified as 'Normal' or 'Stroke' was utilized.
• Genetic Algorithm (GA) was employed for feature selection to reduce redundancy.
• Selected features were fed into InceptionV3, VGG19, and MobileNetV2 deep learning models.

Main Results:

• The integration of GA improved classification accuracy and reduced computational complexity.
• MobileNetV2 achieved the highest accuracy of 97.21% in stroke detection.
• Performance metrics including precision, recall, F1-score, and ROC curves were evaluated.

Conclusions:

• GA-driven feature selection enhances deep learning models for medical image classification.
• MobileNetV2, optimized with GA, shows significant potential for real-time clinical stroke diagnosis.
• This novel approach offers a more efficient and reliable method for stroke detection in emergency settings.