Ensemble Classification of Sleep Apnea Using Single-Lead Electrocardiogram Signal

Sleep Apnea (SA) is a prevalent disorder that impairs sleep quality and can negatively affect one's health. present, sleep centers use polysomnography (PSG) to identify sleep problems. Therefore, there is always space for costeffective and user-friendly techniques. The single-channel electrocardiogram (ECG) signal is used in this study to develop a model for identifying SA episodes. Further, seven different classifiers have been applied and compared to detect SA events using the ECG signal. The learning models were trained and evaluated using performance measures such as accuracy, specificity, F1 score, recall, and precision. Further, a postprocessing step, majority voting, is implemented to enhance the performance of the developed model. The performance of the developed model was also estimated for different volumes of training data. The results depicted the RF classifier as the bestperforming individual classifier, with a classification accuracy of 88.65%. Further, the set of five classifiers (RF, SVM, NN, LDA, NB) with a majority voting post-processing step performed the best with a classification accuracy of 92.55%. The developed system aims to enhance early detection, reduce healthcare costs, and optimize personalized treatment plans to improve sleep disorder diagnostics and patient care.