Sleep Apnea Detection: From Signal Processing to Deep Learning

One of the most widespread sleeping disorders is sleep apnea, which if left untreated, can lead to serious health issues, from hypertension to heart functioning. However, many people are not aware of their condition. The gold standard for identifying sleep apnea is overnight polysomnography at a specialized sleep laboratory. However, because experienced personnel must examine the entire data, these examinations are costly and have a limited number of beds. More patients could be examined if a more rapid diagnosis were made available by an automated detection system. Most autonomous sleep apnea detection algorithms use a collection of human-engineered features, which may miss crucial symptoms. In this study, we present an approach for automatically extracting features from respiratory data and detecting sleep apnea events based on an innovative CNN deep learning model. The algorithm is evaluated on the Apnea-ECG Database, and its results demonstrate sensitivity and accuracy per epoch. This study offers a powerful tool to help medical professionals detect sleep apnea more quickly.