A clinical evaluation of a low-cost strain gauge respiration belt and machine learning to detect sleep apnea

Sleep apnea is a serious and severely under-diagnosed sleep-related respiration disorder characterized by repeated disrupted breathing events during sleep. It diagnosed via polysomnography which an expensive test conducted in sleep lab requiring experts to manually score the recorded data. Since symptoms of are often ambiguous, it difficult for physician decide whether prescribe polysomnography. In this study, we investigate helpful information can be obtained collecting automatically analysing data using smartphone inexpensive strain gauge belt. We evaluate how accurately detect with wide variety machine learning techniques from clinical study 49 overnight recordings. With less than one hour training, distinguish between normal apneic minutes accuracy, sensitivity, specificity 0.7609, 0.7833, 0.7217, respectively. These results achieved even if train only on high-quality entirely separate, clinically certified sensor, has potential substantially reduce cost collection. Data complete night analyzed about second smartphone.