Ordinal Sleep Depth: A Data-Driven Continuous Measurement of Sleep Depth

Conventional sleep staging categorises sleep into discrete stages, which may not capture the continuous nature of sleep depth. We aimed to develop a data-driven continuous measure of sleep depth-ordinal sleep depth (OSD)-using a deep learning framework, and to evaluate its correlation with arousal probability and its association with age, sex, sleep-disordered breathing (SDB) and cognitive impairment. We used 21,787 polysomnography recordings from 18,116 unique patients. A convolutional neural network was trained on 3-s EEG segments to estimate sleep depth continuously, incorporating ordinal regression for the ordered nature of non-REM stages. OSD was compared with the odds ratio product (ORP). Correlations with sleep stages, Arousal Index and clinical variables were assessed. OSD showed a strong linear correlation with arousal probability (Pearson's r = 0.994), slightly outperforming ORP (r = 0.923). Both OSD and ORP reflected expected decreases in sleep depth with advancing age and demonstrated that females have significantly deeper sleep than males across several stages. OSD more accurately captured sleep depth reductions associated with SDB and increasing levels of cognitive impairment, showing significant reductions across all non-REM stages in patients with an increased level of cognitive impairment. OSD as a data-driven measure of sleep depth correlates strongly with arousal probability and effectively captures variations associated with age, sex, SDB and cognitive impairment. The results validate depth as an important dimension of sleep. OSD and ORP provide a nuanced understanding of sleep architecture with physiological and pathological implications.