The Promise of Sleep: A Multi-Sensor Approach for Accurate Sleep Stage Detection Using the Oura Ring

Abstract

Consumer-grade sleep trackers represent a promising tool for large scale studies and health management. However, the potential and limitations of these devices remain less well quantified. Addressing this issue, we aim at providing a comprehensive analysis of the impact of accelerometer, autonomic nervous system (ANS)-mediated peripheral signals, and circadian features for sleep stage detection on a large dataset. Four hundred and forty nights from 106 individuals, for a total of 3444 h of combined polysomnography (PSG) and physiological data from a wearable ring, were acquired. Features were extracted to investigate the relative impact of different data streams on 2-stage (sleep and wake) and 4-stage classification accuracy (light NREM sleep, deep NREM sleep, REM sleep, and wake). Machine learning models were evaluated using a 5-fold cross-validation and a standardized framework for sleep stage classification assessment. Accuracy for 2-stage detection (sleep, wake) was 94% for a simple accelerometer-based model and 96% for a full model that included ANS-derived and circadian features. Accuracy for 4-stage detection was 57% for the accelerometer-based model and 79% when including ANS-derived and circadian features. Combining the compact form factor of a finger ring, multidimensional biometric sensory streams, and machine learning, high accuracy wake-sleep detection and sleep staging can be accomplished.

Deconstructing Commercial Wearable Technology: Contributions toward Accurate and Free-Living Monitoring of Sleep

“…The only resulting physiological trends that can provide discernment on the existence of REM behavior are trends in HR, HRV, and RR that wearables often quantify via PPG. As discussed previously, these measures can demonstrate much variability during REM sleep, unlike other stages; notable spikes in HR and RR can occur just as frequently as mellow lows [ 58 , 70 , 71 , 77 ], suggesting these epochs of the often called “paradoxical sleep” can easily be confused with wake and light sleep, respectively [ 10 ]. Without other data types to ascertain these assumptions that are based primarily on cardiac and respiratory trends, confidently classifying a given epoch as REM is of high difficulty.…”

“…Studies have demonstrated that parameters associated with respiration, heart rate, and movement are most adept at sleep stage recognition [ 78 ]. Specifically, a recent study published by Altini and Kinnunen (2021) demonstrated that device accuracy for four stage detection increased from 57% to 79% when using accelerometer, temperature, HRV, and circadian features rather than just accelerometry alone [ 77 ]. Similarly, studies published by Fonseca et al (2017) and Walch et al (2019) also found sleep algorithm accuracy to increase following the addition of PPG data, as compared to accelerometer data alone [ 70 , 79 ].…”

Sleep and COVID-19. A Case Report of a Mild COVID-19 Patient Monitored by Consumer-Targeted Sleep Wearables

“…It demonstrates that innovative digital tools can have a clinical interest in evaluating patients in real-life conditions. Additionally, it has been shown that data generated by those devices were consistent with those of polysomnography, actigraphy, and self-sleep assessment [ 4 , 5 , 6 ].…”

“…The patient voluntarily recorded his sleep and wake rhythms via three consumer sleep wearables (CSW): Oura ring Gen 2 (Oura) [ 4 ], Fitbit Versa 2 (Fitbit now part of Google) [ 5 ], and iSleep Watch for AppleWatch (iSommeil) [ 6 ]. FitBit Versa 2 watch and iSleep Watch were worn alternatively on the nondominant wrist for 55 days.…”

Using a Homeogram to Detect Sleep in Free-living Animals

“…Efforts to characterize sleep using body motion have, arguably, been more vigorously pursued towards applications related to human health, notably in the past decade due to wearable consumer and medical electronics. This comprises an immense literature concerning accelerometer-based sleep methods-some proprietary-that are tailored towards stereotyped, human sleep phenotypes (Altini & Kinnunen, 2021;Ancoli-Israel et al, 2003;Cole et al, 1992;Enomoto et al, 2009;Nakazaki et al, 2014;Paquet et al, 2007;Sadeh, 2011;Sadeh et al, 1994;Sazonov et al, 2004;van Hees et al, 2018). However, underlying themes emerge in data analysis techniques with regard to sleep.…”