Abstract

Obstructive sleep apnea (OSA) is a sleep related breathing disorder. Identifying severity of airway obstruction is important in OSA severity analysis as well as for treatment success. The apnea hypopnea index (AHI), defined as the total number of full and partial upper airway obstructions per hour, is widely used to diagnose and characterize the severity of OSA. However, recent research shows that AHI provides a crude summary of overnight dynamics of upper airway obstructions. Oesophageal pressure manometry (Pes) is the gold standard method for identifying the severity of individual airway obstruction but, due to the invasive nature, it is less commonly used in sleep laboratories. There is a need for simple automated technology to characterize the severity of airway obstruction. In this work, we propose a method to classify the severity of airway obstruction by analyzing snoring sounds collected through an iPhone 7 smartphone, which requires no physical contact with a subject. For the development of methods, we segmented more than 2000 snoring sound epochs of 5 seconds duration from 7 patients undergoing a polysomnography (PSG) along with Pes. Based on Pes data, we labelled snoring epochs as mild, moderate or severe airway obstruction. We extracted audio features from snoring epochs and used them to train a classifier for multiclass classification. Using 10-fold cross-validation, our methods achieved average accuracy greater than 80% in classifying the severity of airway obstructions. Our results indicate the feasibility of snoring sound in characterizing the severity of airway obstructions. Our non-contact, snoring sound-based technology has the potential to develop into an automatic individual airway obstruction severity analysis system.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call