Comparison of machine learning and non-machine learning methods for the sleep apnea detection using millimeter-wave radar

Abstract

This study investigated the performance of radar-based detection methods for sleep apnea, by comparing a machine learning approach and a non-machine learning approach. We applied a 79-GHz millimeter-wave multiple-input and multiple-output array radar with 12 virtual array elements and performed radar and polysomnography measurements simultaneously to monitor a sleep apnea patient during overnight inspection in a hospital setting. Both radar-based methods successfully estimated the number of apnea events per hour of sleep, with root-mean-square error values of 4.1 and 4.0, indicating that the two methods had comparable accuracy in the radar-based noncontact monitoring of a sleep apnea patient.