Abstracts

Abstract

Objectives: This study assessed the performance of bioradiolocation (BRL) for the determination of sleep and wakefulness in adult population without sleep-disordered breathing (SDB). Methods: BRL is a method for the remote registration of body, limb and organ motions by radar. BRL monitoring and polysomnography (PSG) were performed simultaneously in a sleep laboratory in 32 subjects without SDB (age: 44.2 ± 15.4 years, males/females: 12/ 20, apnea-hypopnea index: 2.43 ± 1.43 episodes/h, sleep efficiency: 77.34 ± 12.53 %). A frequency-modulated continuous-wave radar with quadrature receiver and 8 operating frequencies in the range between 3.6 and 4.0 GHz were used. BRL and PSG records were manually synchronized. Features which represent respiratory activity and body motions were extracted from BRL signal. They were used for the classification of sleep and wakefulness by xgboost classifier. Estimation of classification performance was produced by leave-one-subject-out scheme. Results: The following results were achieved: the accuracy 77.85 ± 6.63 %, Cohen's kappa 0.58 ± 0.16, sleep efficiency error 8.5 ± 7.4 %, total sleep time error 25.9 ± 36.2 min., sleep onset latency error 18.5 ± 34.4 min. Conclusions: These results are consistent with the current state-ofthe- art and demonstrate that BRL is a viable technology for long-term out-patient (home) sleep monitoring.