An Optimization Approach Based on Collective Correlation Coefficient for Biomarker Extraction in the Classification of Alzheimer’s Disease

In this paper, based on Magnetic Resonance Imaging (MRI), the effective biomarkers were efficiently extracted for the classification of Alzheimer’s Disease (AD), mild cognitive impairment (MCI) and health control (HC) with the help of the improved Genetic Algorithm based on the Collective Correlation Coefficient (GA-CCC). Firstly, 544 related features from 68 regions of left and right brain hemispheres were extracted. Secondly, aiming at optimizing Gaussian Process Classifier (GPC) performance, the CCC was employed to help extract the biomarkers for the AD classification and to improve the optimization efficiency of the conventional GA. Finally, experiments showed that the proposed GA-CCC significantly improved the classifications of AD vs. MCI and MCI vs. HC in an efficient way. Plus, many acquired brain regions are known to be strongly involved in the pathophysiological mechanisms of AD.