Multi-Night at-Home Evaluation of Improved Sleep Detection and Classification with a Memory-Enhanced Consumer Sleep Tracker

• DOI: 10.2147/nss.s359789
• Publication Date: 2022-04-13T19:32:07Z
• Authors: Shohreh Ghorbani; Hosein Aghayan Go...; Nicholas IYN Chee; Teck Boon Teo; Andrew Roshan Dicom; Gizem Yilmaz; Ruth LF Leong; Ju Lynn Ong; Michael WL Chee
• Keywords: Sleep; Photoplethysmogram; Sleep Stages; Sleep onset latency

To evaluate the benefits of applying an improved sleep detection and staging algorithm on minimally processed multi-sensor wearable data collected from older generation hardware.58 healthy, East Asian adults aged 23-69 years (M = 37.10, SD 13.03, 32 males), each underwent 3 nights PSG at home, wearing 2nd Generation Oura Rings equipped with additional memory to store raw accelerometer, infra-red photoplethysmography temperature sensors. 2-stage 4-stage classifications using a new machine-learning (Gen3) trained diverse independent dataset were compared existing consumer (Gen2) for whole-night epoch-by-epoch metrics.Gen outperformed its predecessor mean (SD) accuracy 92.6% (0.04), sensitivity 94.9% (0.03), specificity 78.5% (0.11); corresponding 3%, 2.8% 6.2% improvement Gen2 across three nights, Cohen's d values >0.39, t >2.69, p <0.01. Notably, Gen showed robust performance comparable in assessment latency, light sleep, rapid eye movement (REM), wake after onset (WASO) duration. Participants <40 age benefited more upgrade less measurement bias total time (TST), WASO, efficiency those ≥40 years. Males greater improvements TST REM females, while females benefitted deep measures males.These results affirm machine learning training improve device. Importantly, collecting appropriate hardware allows future advancements development or physiology be retrospectively applied enhance value longitudinal studies.