A NOVEL EIGENVECTOR-BASED METHOD TO DETECT MILD ALZHEIMER’S DISEASE USING EVENT-RELAT

Abstract

Event-related potentials (ERPs) are a physiological measure of cognitive function that have shown diagnostic and prognostic utility in Alzheimer’s disease (AD). In this study, we used a novel eigenvector-based technique to better understand brain electrophysiological differences between subjects with mild AD and healthy controls (HC). Using ERPs from 75 subjects with mild AD and 95 HC, we first calculated cognitive task eigenvectors within each subject from three conditions and then calculated second-order eigenvector components to compare the AD group to the HC group. A MANOVA of the three second-level components discriminated between AD and HC multivariately (Wilks’ lambda=.4297, p<0.0001, R2 = .5703), and also on each of the three components univariately (all 3 p-values<0.0001). The eigenvector-based technique used in this study accurately discriminated between the mild AD group and HC. As such, this analysis method adds to our understanding of the differences in ERP signal between AD and HC, and could provide a sensitive biomarker for diagnosis and monitoring of AD progression.