Inspection of short-time resting-state electroencephalogram functional networks in Alzheimer's disease

Functional connectivity has proven useful to characterise electroencephalogram (EEG) activity in Alzheimer's disease (AD). However, most current functional connectivity analyses have been static, disregarding any potential variability of the connectivity with time. In this pilot study, we compute short-time resting state EEG functional connectivity based on the imaginary part of coherency for 12 AD patients and 11 controls. We derive binary unweighted graphs using the cluster-span threshold, an objective binary threshold. For each short-time binary graph, we calculate its local clustering coefficient (Cloc), degree (K), and efficiency (E). The distribution of these graph metrics for each participant is then characterised with four statistical moments: mean, variance, skewness, and kurtosis. The results show significant differences between groups in the mean of K and E, and the kurtosis of Cloc and K. Although not significant when considered alone, the skewness of Cloc is the most frequently selected feature for the discrimination of subject groups. These results suggest that the variability of EEG functional connectivity may convey useful information about AD.