Adaptive Feature Optimization in fMRI Brain Networks for Alzheimer's Disease Detection via GA-SA

This study explores the use of Genetic Algorithm (GA), Simulated Annealing (SA), and a combined GA-SA approach for feature selection in the classification of Alzheimer's Disease (AD) based on fMRI data. The aim is to improve classification model accuracy by selecting the most relevant features through these optimization methods. Experiments were carried out on fMRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), with each test repeated 20 times to ensure stability, and average outcomes reported. The GA-SA method achieved the highest best cost value of 64.6944%, demonstrating its effectiveness in selecting key features, specifically features 1, 2, 5, 9, 10, and 11, which significantly improved the classification accuracy compared to using all features without optimization. The comparison with individual GA and SA approaches, as well as the non-optimized feature set, underscores the critical role of feature selection in boosting model performance. Our results suggest that hybrid optimization techniques like GA-SA are highly effective in enhancing feature selection processes for fMRI-based AD classification.