Sleep Apnea Detection – A Comparative study and efficient solution with modified LeNet-5

Sleep apnea (SA) is a common sleep disease that, if neglected, can have serious cardiovascular and neurological effects. The standard diagnostic method, polysomnography (PSG), involves some professional, multiple electrodes, and cables. Researchers propose using a single-channel signal like electrocardiogram(ECG), easily captured with wearables. However, current ECG-based SA detection methods require expertise. This study suggests employing a modified LeNet-5 convolutional neural network (CNN) to address this challenge. The network considers neighboring segment effects, automatically learning efficient features from training data. Our experiments demonstrate the effectiveness of this approach in SA identification, achieving results comparable or superior to traditional machine learning methods.