Discriminating early stage AD patients from healthy controls using synchronization analysis of EEG

In this paper we study how the meso-scale and micro-scale electroencephalography (EEG) synchronization measures can be used for discriminating patients suffering from Alzheimer's disease (AD) from normal control subjects. To this end, two synchronization measures, namely power spectral density and multivariate phase synchronization, are considered and the topography of the changes in patients vs. Controls is shown. The AD patients showed increased power spectral density in the frontal area in theta band and widespread decrease in the higher frequency bands. It was also characterized with decreased multivariate phase synchronization in the left fronto-temporal and medial regions, which was consistent across all frequency bands. A region of interest was selected based on these maps and the average of the power spectral density and phase synchrony was obtained in these regions. These two quantities were then used as features for classification of the subjects into patients' and controls' groups. Our analysis showed that the theta band can be a marker for discriminating AD patients from normal controls, where a simple linear discriminant resulted in 83% classification precision.