Fuzzy set training for sleep apnea classification

Untreated sleep apnea may result in higher risk for daytime drowsiness, heart conditions, high blood pressure, stroke, reduced cognitive function, and development of dementia. Polysomnography is the primary diagnostic tool for sleep apnea, but the cost and uncomfortable environment hinder diagnosis in many individuals. Home-based sleep monitoring systems would increase accessibility and comfort, but require robust signal analysis for quality analysis. Machine learning has been applied for sleep apnea classification, but more improvements would enhance effectiveness. In this study a Fuzzy set system with tuning was developed for classification of sleep apnea or hypopnea events during sleep for individuals at risk of sleep apnea. Annotation files from a sleep study available on Physionet database were analyzed for classification of sleep disruption events. The performance of the developed fuzzy set algorithm was compared with classification by other machine learning algorithms using the Weka system. Improved algorithms for classification of sleep events would be useful toward development of sleep monitoring systems that potentially would encourage individuals with sleep events to seek treatment.