Automatic Obstructive Sleep Apnea Identification Using First Order Statistics Features of Electrocardiogram and Machine Learning

Obstructive Sleep Apnea (OSA), or cessation of breathing during sleep, is caused by a blocked upper airway. OSA is one of the causes of sudden death caused by heart failure during sleep. Given the impact of OSA, early detection is therefore necessary. This article presents an innovation in OSA detection using first-order statistical features of an electrocardiogram (ECG) and machine learning. The first-order statistical features were obtained from the RR interval of the ECG single-lead. Various statistical features were evaluated for OSA detection. Moreover, machine learning techniques such as Linear Discriminant Analysis (LDA), Artificial Neural Networks (ANN), K-Nearest Neighbors (K-NN), and Support Vector Machines (SVM) were investigated for OSA detection. The accuracy, sensitivity, and accuracy values of each classification are the performance outcomes of machine learning. The proposed OSA detection method was validated using clinical data from patients diagnosed with sleep apnea. When using KNN for OSA detection, the results showed an accuracy of 83.91%, sensitivity of 92.31%, and specificity of 67.19%. Using SVM, the OSA detection demonstrated an accuracy of 83.75%, sensitivity of 92.97%, and specificity of 65.80%. The ANN method for OSA detection yielded an accuracy of 80.83%, sensitivity of 92.67%, and specificity of 59.25%. Lastly, OSA detection using LDA exhibited an accuracy of 79.27%, sensitivity of 82.27%, and specificity of 73.29%. From the results, it was concluded that the KNN classification, with all its features, provided the best performance for OSA detection, achieving the highest accuracy of 83.91%.