Deep learning for automated sleep staging using instantaneous heart rate

• DOI: 10.1038/s41746-020-0291-x
• Publication Date: 2020-08-20T10:07:20Z
• Authors: Niranjan Sridhar; Ali Shoeb; Philip Stephens; Alaa Kharbouch; David Ben Shimol; Joshua Burkart; Atiyeh Ghoreyshi; Lance Myers
• Keywords: Sleep

Abstract Clinical sleep evaluations currently require multimodal data collection and manual review by human experts, making them expensive unsuitable for longer term studies. Sleep staging using cardiac rhythm is an active area of research because it can be measured much more easily a wide variety both medical consumer-grade devices. In this study, we applied deep learning methods to create algorithm automated stage scoring the instantaneous heart rate (IHR) time series extracted from electrocardiogram (ECG). We trained validated on over 10,000 nights Heart Health Study (SHHS) Multi-Ethnic Atherosclerosis (MESA). The has overall performance 0.77 accuracy 0.66 kappa against reference stages held-out portion SHHS dataset classifying every 30 s into four classes: wake, light sleep, rapid eye movement (REM). Moreover, demonstrate that generalizes well independent 993 subjects labeled American Academy Medicine (AASM) licensed clinical staff at Massachusetts General Hospital was not used training or validation. Finally, predicted our reproduce previous studies correlating with comorbidities such as apnea hypertension demographics age gender.