Project Description:
Sleep apnea is a sleep disorder characterized by interruptions in breathing while asleep. It is traditionally diagnosed through a clinical polysomnography (PSG) sleep study, but this is highly invasive, costly, and geographically limiting. Photoplethysmography (PPG) is a promising alternative that can be obtained through wearables such as smart watches and fitness trackers. We investigate several machine learning models to detect sleep apnea events from PPG signals including traditional models and deep learning techniques. From these predicted events, we compute a patient’s apnea-hypopnea index (AHI), a score of the severity of sleep apnea. We also developed a web application for demonstrating our models on sample PPG signals.
