The Effects of Anthropometric Parameters on the Breathing Sound Features while Screening Obstructive Sleep Apnea during Wakefulness

Abstract

Anthropometric characteristics, such as gender, body mass index (BMI), age, etc. are considered as risk factors for obstructive sleep apnea (OSA). These information are used for screening OSA during wakefulness, but they provide a poor specificity compared to our screening method using tracheal breathing sound analysis. Despite that, one of the main challenges of using breathing sounds analysis for classification of OSA during wakefulness is the effect of confounding variables. Breathing sounds are not only affected by OSA, but also by the anthropometric factors. In this work, we investigated which sound features show the least correlation to anthropometric factors. Tracheal breathing sounds of 114 individuals (66 subjects with apnea/hypopnea index (AHI)<15 and 48 with AHI>15) were recorded during wakefulness in supine position. Spectra and bi-spectra of the signals of the two AHI groups were analyzed to extract the most significant features. Our results suggest it is possible to find the best features with high sensitivity to AHI and least sensitivity to confounding variables.