A Machine Learning-Based Models For Intelligent Automated Sleep Staging Classification System Using Polysomnography Data

It is essential in clinical practice to consider an automated tool for identifying the sleep stages” irregularities and for accurate analysis of sleep quality. Sleep staging is essential for the diagnosis of different sleep-related disorders. Single-channel electroencephalography (EEG) is frequently used in sleep staging studies since it complies with recommended practices for sleep staging. This work's main objective is to automatically analyze sleep patterns, which helps to enhance the knowledge of the boundaries between the sleep stages by extracting the multi-scale temporal context using polysomnography (PSG) signals. The extracted features from the recorded calls are fused and input into the different machine learning models. The whole experiment was performed on the Physionet Sleep-EDF database. The Random Forest (RF) classifier performed better than the K-Nearest Neighbor (KNN) classifier, achieving more accurate classification performance results for five-class sleep stages classification problems. The experimental findings demonstrate that the RF classifier, which captures variations in sleep behavior characteristics over distinct sleep stages, outperforms state-of-the-art approaches in several assessment criteria.