Obstructive Sleep Apnea Detection using Fuzzy Approximate Entropy of Extrema based on Multiple Moving Averages

Abstract

Obstructive sleep apnea (OSA) is a common upper respiratory tract disease, which is related to autonomic nervous system (ANS) dysfunction and associated with reduced heart rate variability (HRV). Fuzzy approximate entropy of extrema based on multiple moving averages (Emma-fApEn) can effectively analyze the physiological sympathetic tone in a short period of time during sleep. In this study, we compared fApEn-minima and fApEn-maxima obtained with Emma-fApEn with classic time-frequency domain indices using electrocardiogram(ECG) recordings from the PhysioNet database. The empirical results showed that Mean and LH could significantly differentiate OSA recordings from healthy recordings. Compared with support vector machine (SVM) and k-nearest neighbor classification (KNN), random forest (RF) provided the highest accuracy in OSA detection. Therefore, Emma-fApEn could analyze the decrease in the complexity of sympathetic tone in OSA patients during sleep.