A Model of Alzheimer's disease and mild cognitive impairment based on EEG coherence

EEG coherence was used to model the integrity of neural connectivity in patients with Alzheimer's Disease (AD), Mild Cognitive Impairment (MCI), and age-matched controls. Our model was derived from the EEG data of 16 AD patients, 24 MCI patients and 16 age-matched controls. We computed the coherence of the relevant channel pairs believed to represent fiber pathways. Coherence values were compared across groups of AD patients, MCI patients, and controls, and the Kruskal-Wallis test was used to determine those channel pairs that exhibited significant differences between MCI or AD patients and controls. Because increased and decreased EEG coherence (relative to controls) has been found to be associated with fiber damage, channel pairs showing significant differences in EEG coherence were rated as damaged. The data show that while AD patients have more damaged pathways than MCI patients, the areas of damage are similar for both groups. The significance of this model is its future application (perhaps in conjunction with white matter imaging techniques) to the development of connectivity models of cognitive decline in AD and MCI patients.

I. Introduction

THE magnitude squared coherence function, or simply “coherence”, is a measure of synchrony between two independent signals. At any given frequency in Hz, the coherence between two independent signals and is defined by (1) [1], where is the cross-spectral density of and y, and and are the power spectral densities of and respectively. $$C_{xy}(f)={\bigg\vert P_{xy}(f)\bigg\vert ^{2}\over P_{xx}(f)P_{yy}(f)}. \eqno{\hbox{(1)}}$$