0096 Sleep Apnea Diagnosis using Tracheal Signals and Oximetry

Abstract

Introduction Diagnosing sleep apnea requires detection of apneas and hypopneas during sleep either via in-laboratory overnight polysomnography (PSG) or portable in-home sleep apnea testing (HSAT). While PSG is the optimal method, it is expensive, inconvenient and often inaccessible for patients. Although HSATs are more convenient and less expensive than PSG, they are not as accurate and have relatively high failure rates because of the nature of the sensors used to measure respiratory variables. We have developed a HSAT, the Patch, that is unique in that it is very simple and reliable because it measures respiratory variables using tracheal motion and sound combined with oximetry to detect respiratory events. In this study we tested the ability of the Patch to detect respiratory events versus simultaneous PSG. Methods Participants were adults with a suspected sleep disorder referred to the sleep laboratory at Toronto Rehabilitation Institute for PSG. Simultaneous to the PSG, the Patch, consisting of a module containing a microphone and a 3-D accelerometer that was affixed to participants’ suprasternal notch, and a finger oximeter. After filtering the tracheal signals, the envelope of the tracheal motions in the cranial and postero-anterior directions and tracheal sound envelope were extracted. Along with tracheal features, the amplitude and slopes of oxygen desaturations were also extracted and, were fed into a supervised deep neural network model to detect apneas and hypopneas. The total number of detected events was divided by total estimated sleep time to estimate the apnea-hypopnea index (AHI). The performance of the model in diagnosing sleep apnea was evaluated by sensitivity (AHI≥15) and specificity (AHI<15). The relationship between the estimated AHI and PSG-based AHI was quantified using Pearson correlation. Results Ninety-nine participants (42 females, age: 48±16 years, body mass index: 29.2±5.2 kg/m2, and AHI: 15.8±19.4 events/hour) completed the study. We found that the Patch had 88.6% sensitivity and 89.1% specificity for diagnosing sleep apnea. Strong agreement was observed between the estimated and reference AHI values (r = 0.92, p < 0.001). Conclusion The Patch is a novel, robust and convenient portable device that provides an accurate means of detecting and quantifying sleep apnea. It has the potential to provide reliable home-based sleep apnea monitoring. Support (If Any) Funded by Bresotec Inc.