Simultaneous estimation of sleep duration and sleep apnea-hypopnea using millimeter-wave radar

Comorbid insomnia and sleep apnea is increasingly recognized, yet detection remains challenging due to limitations of contact-based monitoring such as polysomnography (PSG) and wearable devices. Millimeter-wave radar (MWR) may provide a non-contact alternative for simultaneous assessment of sleep-disordered breathing and sleep duration. This study evaluated the accuracy of MWR in estimating the respiratory event index (REI) and sleep duration, and identified optimal post-movement exclusion windows to enhance concordance with PSG standards. Sixty-two adults underwent overnight PSG with concurrent MWR recording. REI and estimated sleep duration (ESD) were calculated after excluding data within 1, 2, or 3 min following large body movements. Agreement between REI and PSG-derived apnea-hypopnea index (AHI), and between ESD and PSG-derived total sleep time (TST), was assessed using Bland-Altman analysis. The 3-min exclusion window yielded the smallest bias and narrowest limits of agreement for REI (mean bias = + 0.1 events/h; 95% confidence interval = - 8.2 to 8.4), matching PSG-derived AHI (18.9 ± 16.9 vs. 19.0 ± 14.4 events/h). In contrast, the 1-min exclusion window provided the best agreement for sleep duration (mean bias = - 26 min; 95% confidence interval = - 72 to 20), aligning with PSG-derived TST (422 ± 80 vs. 396 ± 71 min). MWR enables simultaneous, non-contact estimation of sleep-disordered breathing and sleep duration with clinically acceptable concordance to PSG. Metric-specific optimization-3 min for REI and 1 min for ESD-supports MWR as a low-burden tool for screening and monitoring in clinical and home environments.