Machine Learning Approach for Early Diagnosis of Alzheimer's Disease Using rs-fMRI and Metaheuristic Optimization with Functional Connectivity Matrices

Abstract

Alzheimer's disease is a cognitive disease that occurs with memory loss and usually occurs in older ages. Nowadays, considering that the risk of Alzheimer's increases every year as the life expectancy of the individual increases and has become a global problem, the importance of early detection of Alzheimer's disease is emphasized. This study uses brain connectivity matrices with informative features obtained from resting state functional MRI (rs-fMRI) data as features for early and effective diagnosis of Alzheimer's. Feature selection is performed using SMA and PSO algorithms, both of which create a machine learning model using optimization algorithms for Alzheimer's diagnosis. Among the two different models, the SMA-based model achieved approximately 92% success in diagnosing Alzheimer's disease with an average of 104 features. This revealed that the proposed method has good potential.