Abstract Introduction Wearable technology offers the potential for accessible monitoring for sleep disorders. The SleepImage ring (SI, MyCardio, Denver) acquires actigraphy and oximetry/photoplethysmography data processed via proprietary cardiopulmonary coupling analysis to identify sleep and respiratory events. While there are limited data validating oximetry measurements, the validation of SI sleep analysis versus conventional polysomnographic (PSG) sleep staging has not been published. Our objective was to evaluate the accuracy of SI ring sleep-wake and respiratory data compared with simultaneous in-laboratory complete PSG. Methods Patients with suspected OSA underwent overnight in-laboratory PSG while wearing the SI ring. SI-generated sleep and respiratory metrics were compared with psg scored manually using AASM3.0 criteria. Agreement between PSG and SI was assessed using paired comparisons (t-test), intraclass correlation coefficients (ICC) and Bland-Altman (B-A) analysis; we also performed an epoch-based concordance analysis. Results To date, we have studied 15 patients (45.1 ± 13.8 y; 12 males; BMI 32.1 ± 6.9 kg/m²). Sleep-wake epoch concordance data (%, (# epochs)) are shown in the Table below. SI overestimated total sleep time (343.7 ± 87.7 min vs 379.4 ± 52.3, PSG vs SI, respectively, p=NS; B-A bias PSG vs SI = -35.7 min (95% CI -180.2 - 144.5); ICC =0.18). SI underestimated wake time (96.4 ± 95.3 vs 48.2 ± 29.9 min, p = 0.04; B-A bias = 48.2 (-163.3 - 211.4), ICC=0.26). Mean NREM duration was comparable (292.4 ± 75.4 vs 289.1 ± 48.7 min, p=NS, B-A bias = 3.3 (-138.3 - 141.50), ICC=0.40), while SI substantially overestimated REM duration (51.3 ± 32.6 vs 90.3 ± 30.7 min, p 0.01; B-A bias = -39.0 (-70.8 - 31.8), ICC =0.20). Measures of oxygenation showed close agreement, including mean and minimum SpO2, T90, 4% ODI % and 3% ODI (21.0 ± 23.5 vs 19.0 ± 23.0 events/h, p=NS; B-A bias = 2.0 (-18.9 - 20.9), ICC 0.92). However, SI underestimated AHI (33.5 ± 28.3 vs 21.0 ± 20.6 events/h, p = 0.005, B-A bias = 12.5 (-29.1 - 40.6), ICC=0.75) and RDI (36.5 ± 26.9 vs 24.8 ± 19.8 events/h, p = 0.009, B-A bias = 11.7 (-29.2 - 40.9), ICC=0.72). Conclusion Our findings to date indicate that SI overestimates sleep time, which appears largely related to overidentification of REM at the expense of wakefulness, while NREM sleep is more reliably identified. SI oxygenation indices correlate very closely with PSG while AHI and RDI are somewhat underestimated. These differences have potential implications for patient management. This abstract is funded by: Sleep Image rings and report data were provided without charge by the manufacturer.No other funding was received for this study.
Alsayari et al. (Fri,) studied this question.