Advanced Deep Learning Model for Sleep Stages Classification Using Single Channel EEG

Abstract

Sleep stages scoring is a first method to identify whether a subject is affected with any disorder or not. Scoring of sleep stages done through the analysis of Polysomnogram (PSG). Sleep stage ordering assumes a significant job in the conclusion of sleep-related illnesses as a rule performed via skilled experts utilizing visual investigation of brain wave records from the scalp of the subject. PSG signal incorporates Electroencephalogram (EEG), Electrocardiogram (ECG), and Electromyogram (EMG) signals, in which consider single-channel EEG signal only to predict the sleep stages depends on the supervised learning of 5 stage scoring. With the rapid development in this Artificial Intelligence (AI) domain, nowadays classification of sleep stages done by automation is very simple by using Deep Learning models (DL). In this work we considered two best popular algorithms such as Convolution Neural Networks (CNN) and Long-Short Term Memory (LSTM). The proposed method introduces an extensible deep learning network approach to score sleep stages from raw PSG signals. Proposed model trained and tested with the most popular database taken from physio bank such as Sleep EDFX. The proposedmodel yields an accuracy of 89 % for CNN and 84 % for LSTM.