Detection of Breathing Related Sleep Disorders using Audio Signals

Sleep disordered breathing, including obstructive sleep apnea (OSA), is a prevalent group of phenomena causing many health problems such as daytime fatigue and systemic hypertension. The gold standard for OSA diagnosis is polysomnography (PSG), which has several drawbacks like unnatural sleep for the patient and high cost accompanied with long waiting lists. We hypothesize that acquisition of the sleep audio signal in a non-contact way and processing this signal is an efficient method that can lead to a reliable and low-cost screening tool. The proposed OSA detection system concludes a clock sound detector sub-system which was combined with preliminary research done in our lab. It achieved detection accuracy of 96.5% when validated using 85 home-recorded subjects. The overall system achieved correlation of R=0.62 (p-value 9E-11) between AHI estimation and the AHI estimated by a WatchPat device; the diagnostic agreement was 69.32%. This kind of system will enable the patient to undergo the test by himself, in his own house, and get a reliable OSA assessment in the next morning.