A New EEG Marker of Cognitive Decline and Memory Performance for the Elderly

This study develops new markers of the Alzheimer''s disease and other types of dementia for the elderly based on the all-night EEG of a single channel. The sleep EEG is transformed into 5 power time series corresponding to the delta, theta, alpha, sigma and beta frequency bands. Then, the mutual information and wavelet coherence are adopted to develop a measurement for the similarity between the power time series of a frequency band pair, denoted as the half phase variance (PV). The PV values are used as markers of the Alzheimer''s disease and other types of dementia for the elderly.

First of all, the sleep EEG from electrodes C3-A2, C4-A1, O1-A2 and O2-A1 have been recorded for 7 dementia''s patients (including AD patients and non-AD type patients), 19 age-matched normal controls, and 5 normal young people as baseline in our algorithm. It is found that the PV of theta and alpha band power time series prominently increases in dementia''s patients as compared with the controls in the C3 and O1 channels with p-value<0.001, especially for AD patients. In addition, it also increases significantly in dementia''s patients in the C4 channel with p-value<0.01. These PV values seem to be able to reflect neurophysiological degeneration and thus may serve as a marker to identify dementias. Compared with conventional approach, this method is advantageous because only one channel measurement is required.

Furthermore, the results of cross-similarity defined as previous studies have also been shown in the study. The significant difference of PV values only appears in the band pair of alpha of the left local posterior (C3-A2 & O1-A2) that the PV values are slightly higher in dementias than in controls (p<0.05) and the right local posterior (C4-A1 & O2-A1) have shown the same results. Thus, for the bilateral parietal-occipital regions, the PV values of alpha band pair might be another maker to identify dementias.

Based on above markers, the relationships between PV of certain band pairs and cognitive performance are assessed as well. The results demonstrate that the PV values are strongly correlated with MMSE scores. In addition, the relations with the subtests of WMS scores and the PV values are also quantitative correlated. These results demonstrate our proposed method seems to be able to effectively link to well-documented neuropsychological tests. Thus, it indicates that the proposed PV possess a superior extension for further research.

Lastly, the age-related changes of PV are considered to try to figure out the age effects on the EEG similarity. For cross-similarity, the PV values of the same frequency band present the lower value commonly in young people and the higher values commonly in patients. Generally, the distributions of PV from the young, the normal elderly to dementias of above cases approach the positive relations. For the auto-similarity of C3 and C4 channels, the PV values of alpha and theta band-power time series also shows the same result. These consistent trends can be concluded that some sources of pathological reason might be similar between dementias and aging.