Gender dependant snore sound based multi feature obstructive sleep apnea screening method

Obstructive Sleep Apnea (OSA) is a serious sleep disorder that occurs due to collapsing upper airways (UA). More than 80% of OSA sufferers remain undiagnosed and the situation demands simplified, convenient technology for community screening. Almost all OSA patients snore and snoring is the earliest nocturnal symptom of OSA. Snore signals (SS) are produced due to vibration of soft tissues in the narrowed parts of the UA. It is known that the UA properties are gender specific. In this paper, we work under the hypothesis that gender specific analysis of snore sounds should lead to a higher OSA detection performance. We propose a snore based multi-parametric OSA screening technique, which incorporates the gender differences in the algorithm. The multi feature vector was modeled using logistic regression based algorithms to classify subjects into OSA/non-OSA classes. The performance of the proposed method was evaluated by carrying out K-fold cross validation. This procedure was applied to male (n=51) and female (n=36) data sets recorded in a clinical sleep laboratory. Each data set consisted of sound recordings of 6-8 hr. duration. The performance of the method was evaluated against the standard laboratory method of diagnosis known as polysomnography. Our gender-specific technique resulted in a sensitivity of 93±9% with specificity 89±7% for females and sensitivity of 91±8% with specificity 89±12% for males. These results establish the possibility of developing cheap, convenient, non-contact and an unattended OSA screening technique.