Power spectral density and high order bispectral analysis of Alzheimer's EEG

Abstract

In this paper, power spectrum density (PSD) and high order bispectral (BIS) analysis are proposed to investigate the abnormalities of electroencephalogram (EEG) signals which are collected from fourteen Alzheimer's disease (AD) patients and fourteen age-matched normal controls. The power spectrum density (PSD) estimated by AR model is first used to evaluate the frequency distribution in AD brain. It is found that for the whole frequency band, the significant alteration occurs at the electrode FP1, FP2, O1 and T3, and for the sub-bands, the relative PSD shows significant increase in the theta frequency band and decrease in the alpha 2 frequency bands in AD. In order to explore the nonlinear phase-coupling information, high order bispectral analysis is further applied to O1 electrode in the whole frequency band. It is demonstrated that the non-Gaussianity and nonlinearity of EEG are decreased in AD by bispectral analysis. The obtained results show that analysis of PSD and BIS can be taken as a potential comprehensive measure to distinguish AD patients from the normal, which may benefit our understanding of the disease.