Automatic classification of apnea and normal subjects using new features extracted from HRV and ECG-derived respiration signals

Abstract A novel framework for automatic detection of obstructive sleep apnea (OSA) is introduced in which a symbolic dynamics method, alphabet entropy, along with other well-known features such as fuzzy/approximate and sample entropy are calculated from ECG-derived respiration (EDR) and heart rate variability (HRV) signals. In addition, six different algorithms are employed in the extraction of the EDR signal from a single-lead ECG, and the results are compared. The sequential feature selection method is applied to pick the most effective features. Finally, the picked features are fed into the different classifiers to classify OSA patients and normal subjects. The Physionet Apnea-ECG and Fantasia datasets are utilized to assess the proposed OSA detection method and EDR extraction algorithms, respectively. The results show that the GentleBoost classifier has achieved the accuracy of 93.26% and 100% in per-segment and per-recording classifications, respectively. The proposed automatic OSA detection system outperforms other existing state-of-the-art methods in per-segment classification.