Optimization Algorithms for Brain MRI-Based Alzheimer’s Disease Classification: A Comprehensive Review and Methodological Framework

Abstract

Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder that significantly affects cognitive functions. Magnetic Resonance Imaging (MRI) functions as a non-invasive diagnostic method which scientists use to identify Alzheimer's disease at its first stage. Deep learning models including Convolutional Neural Networks (CNNs), and Transformer-based architectures have emerged as the leading technology for AD classification during the last several years. The performance of neural networks depends on the correct selection and adjustment of hyperparameters and weights, and network structure design. The paper provides a complete analysis of traditional and nature-inspired and contemporary meta-heuristic optimization methods which serve AD classification purposes. We propose a methodological framework which combines deep learning with advanced optimizers including AdamW, Lookahead, Bayesian Optimization, Particle Swarm Optimization (PSO), Genetic Algorithms (GA), and recent hybrid strategies. The paper presents a summary of optimization-based AD classification methods which identifies their current limitations and future research directions.