Key points are not available for this paper at this time.
Abstract Introduction Polysomnography (PSG) requires on-body sensors for respiratory effort measurement, that can adversely affect sleep quality and accuracy of detection of sleep related breathing disorders (SRBD). We developed a non-contact 60 Hz pulsed wave radar system with integrated advanced signal processing to collect respiratory effort, heart rate and gross body movements data using phase differencing to a sub-mm resolution. In this study we tested the performance of this system in detecting SRBD events against standard PSG in human participants. Methods During participants’ clinically indicated PSG, we collected concurrent data from the radar system placed under the bed on a tripod mount. Algorithms were developed to quantify breath amplitude reductions collected by radar and compared to the time-synced signals derived from respiratory impedance plethysmography belts. A Convolutional Neural Network (CNN) was trained on radar in-phase/quadrature (I/Q) data for each 30 second epoch to output probability of apnea and hypopnea events using standard event definitions. Receiver Operating Characteristics (ROC) plot and confusion matrix for the trained network were generated to provide accuracy of event predictions. Results A total of 60 participants completed the study. The participants had a mean age of 65 years (Inter Quartile Range 13.5 years), and 56% subjects were male. The mean AHI for the entire cohort was 12.9 (IQR 14.75) events/hour. ROC plots showed an Area Under the Curve (AUC) of 0.98 for events detection by radar compared to human scored PSGs. The confusion matrix was generated at an operating point of 95% sensitivity and 80% specificity. Conclusion The non-contact high frequency radar system showed excellent performance in respiratory event detection in patients with sleep-related breathing disorders in comparison to the gold standard PSG system. Use of this technology is expected to reduce the number of body contact sensors used during conventional sleep studies and thus improve patient comfort and accuracy of results. Support (if any) Research reported in this abstract was supported by National Institute of Aging of the NIH under award number: R44AG060779
Goswami et al. (Sat,) studied this question.