Sleep Breathing Disorders Detection with Bioradar Using a Long Short-Term Memory Network

Development of effective non-contact ways for long-term sleep respiratory-related sleep disorders detection, which may indicate the presence of different health and lifethreatening conditions, is an up-to-date task of sleep medicine. The paper presents a device for remote long-term sleep respiration pattern monitoring based on the analysis of a bioradar signal and processing algorithm for detection respiratory-related sleep disorders. The method was validated utilizing data of 15 volunteers, which underwent a sleep study in a sleep laboratory of Almazov National Medical Research Centre. The proposed method is based on the usage of a long short-term memory network to detect breathing disorders during sleep. We achieved accuracy and Cohen's kappa of 0.97 and 0.80 for respiratory-related sleep disorders classification, respectively. The results might be used while creating new methods for remote detection of sleep movement disorders.