Estimation of Circadian Rhythms Using Complexity Analysis with Temporal Scale Dependency in Electroencephalogram Signals

Disturbances in circadian rhythms have been recently associated with a variety of healthy states and psychiatric pathologies. Therefore, estimating the degree of circadian rhythm disturbance is important for discriminating psychiatric disorders from healthy conditions. Electroencephalogram (EEG) allows to detect brain activity directly, but the recorded signal combines neural activity across multiple time scales. The complexity of brain activity across multiple time scales has been previously quantified using multiscale entropy (MSE) analysis. In this study, we investigated whether MSE analysis of EEG data can detect circadian rhythms. Our results show that in the daytime, the complexity of brain activity is increased at larger temporal scale, and that MSE analysis detects these changes more accurately than conventional power analysis. Because complexity at large temporal scales arises from the long-range connectivity in brain networks, we suggest that the decrease in this EEG pattern complexity by night is mediated by melatonin, which suppresses neural firing and reduces wide-range interactions between brain regions. Our method can be applied for the EEG-based analysis of circadian rhythms in longitudinal studies and may help to diagnose healthy states and psychiatric conditions.