A Quantitative Representation of Continuous Brain State During Sleep

Abstract

Electroencephalography (EEG) recording during sleep is a powerful tool for studying neural oscillatory dynamics. Current practice, however, limits our ability to observe the potentially-important features of the sleeping brain because it focuses on discrete stage-based averaging of manually labeled metrics. This study proposes a high-resolution objective framework that can comprehensively represent individualized continuous brain states during sleep. We develop a computational framework around a novel visualization called the slow-oscillation power spectrogram (SOPS), which describes EEG spectral activity as a function of a continuum of sleep depth rather than over discrete stages. Continuous sleep depth is represented by the instantaneous ratio of slow-oscillation (0.5 - 2 Hz) power to the total power (SO-power ratio). We analyzed sleep EEG of two consecutive nights of polysomnography obtained from 10 healthy subjects (mean age of 25.8 +/- 5.09 years, 50% female). Through visualization and quantitative analysis, our results suggest that the SOPS can offer a concise high-resolution summary of an individual's unique depth-of-sleep profile. Our framework provides a basis for exploring the critical biomarker of neurocognitive functions for which sleep EEG properties are significantly relevant.