Parameters Analysis of Sample Entropy, Permutation Entropy and Permutation Ratio Entropy for RR Interval Time Series

Abstract

Proper parameters can improve performance of entropy methods for discerning electrocardiogram (ECG) signals. So, we tried to determine proper parameters of three entropy methods i.e., a novel permutation ratio entropy (PRE), sample entropy (SmpE) and permutation entropy (PE) for discerning several typical ECG RR interval recordings i.e., normal sinus rhythm (NSR), congestive heart failure (CHF) and NSR and arrhythmia RR (ARR) interval recordings. The three entropy methods were first calculated for a logistic sequence to evaluate their sensitivity to dynamic property changes within a time series. Their capabilities of distinguishing complexity between NSR and CHF, NSR and ARR, and CHF and ARR RR interval recordings were compared. Statistical differences between the three entropy values for normal (i.e., NSR) and abnormal RR interval recordings (i.e., CHF and ARR) were analysed respectively. Performance of the entropy methods in simultaneously discerning the three groups (i.e., NSR, CHF and ARR groups) were also compared. PRE more accurately reflected logistic sequence changes from period doublings to chaos than SmpE or PE did. In experiments with real data, PRE correctly yielded higher values on NSR RR recordings than on CHF and ARR recordings and exhibited significant differences (p < 0.01) on more parameter pairs than SmpE and PE did.