Feasibility Study for Apnea Screening in Patients' Homes Using Radar and Machine Learning Method

Sleep apnea is a respiratory disorder that occurs during sleep with shallow or interrupted breathing, which negatively affects sleep quality. In clinical practice, polysomnography (PSG) is a gold standard method for diagnosing apnea, by directly applying multiple wired sensors to the patient's body to acquire physiological signals such as respiration, oxygen saturation, brain wave, heart rate, body posture, and others. However, PSG can only be completed overnight in hospital sleep centers and under the supervision of a licensed technician. PSG is not readily available, expensive, time-consuming, and cannot provide long-term monitoring. Many sensors attached to the bodies of patients also induce discomfort. To solve this problem, a home-based, efficient, affordable, and non-contact method is required for the rapid detection of apnea. We proposed a novel non-invasive and non-contact sensing system and conducted a study to evaluate the feasibility of the home system as a rapid test for the diagnosis of apnea. Millimeter wave (mmWave) radar and machine learning methods were adopted to implement such a system. Clinical data from 100 individuals diagnosed with suspicious sleep apnea were collected from recordings obtained in the sleep center of the participating hospital and in their homes. The experimental results revealed that apnea was detected with an accuracy rate of 93% using a mmWave radar. The results of the study may inspire the next research and development of a non-invasive and non-contact sensing for the detection of apnea.