EEG classification of mild and severe Alzheimer's disease using parallel factor analysis method: PARAFAC decomposition of spectral-spatial characteristics of EEG time series

Abstract

Electroencephalograms (EEG) recordings are now widely used more and more as a method to assess the susceptibility to Alzheimer's disease. In this study, we aimed at classifying control subjects from subjects with mild cognitive impairment (MCI) and from Alzheimer's disease (AD). For each subject, we computed the relative Fourier power of five frequency bands. Then for each frequency band, we estimated the mean power of five brain regions: frontal, left temporal, central, right temporal and posterior. There were an equivalent number of electrodes in each of the five regions. This grouping is very useful in normalizing the regional repartition of the information. We can form a three-way tensor, which is the Fourier power by frequency band and by brain region for each subject. From this tensor, we extracted characteristic filters for the classification of subjects using linear and nonlinear classifiers. © 2009 Springer Netherlands.