Aperiodic-Aware EEG Features for Cross-Subject Acute Sleep Deprivation Detection

Acute sleep deprivation impairs vigilance and cognitive stability. A potential contender is resting-state electroencephalography (EEG), though traditional features of bandpower will be potentially confounded by large band aperiodic activity. In this study, the authors assessed the use of aperiodic-aware EEG characteristics as a useful indicator of detecting acute sleep deprivation among the subjects. EEGs of an eyes-open resting-state recorded at the OpenNeuro ds004902 dataset were analysed. In-subject, when statistical Tests demonstrated that the aperiodic exponent amplified in sleep deprivation in both reduced-channel and all-channel schemes, and that this trend existed in an inflexible sensitivity sub-group. There was no significant relationship between session order and age and the primary exponent change. By contrast, cross-subject discrimination with lightweight classifiers was only moderate, implying that the generalization process is still difficult despite the repeatability of a physiological change. These results give a feasible criterion upon which future EEG-based monitoring analyses may rely.