Comparison of Time-Domain, Frequency-Domain and Non-Linear Analysis for Distinguishing Congestive Heart Failure Patients from Normal Sinus Rhythm Subjects
Yan Wang,Shoushui Wei,Shuai Zhang,Yatao Zhang,Lina Zhao,Chengyu Liu,Alan Murray
DOI: https://doi.org/10.1016/j.bspc.2018.01.001
IF: 5.1
2018-01-01
Biomedical Signal Processing and Control
Abstract:It is known that patients with congestive heart failure have reduced ability to modulate heart rate in comparison with normal subjects. However, the characteristics of these changes is not well understood. This study therefore investigated the characteristic features of heart rate changes to assess how they differed between both groups. Fifty-two normal sinus rhythm subjects and 18 congestive heart failure patients from the PhysioNet database were studied. Nine common heart rate indices were studied: three time-domain indices (MEAN RR interval, standard deviation of successive RR SDNN, and square root of mean squared differences of successive RR RMSSD), three frequency-domain indices (normalized low frequency power LFn normalized high-frequency power HFn and their ratio LF/HF), and three non-linear indices (vector length index VLI, vector angle index VAI and sample entropy SampEn). Two 5-min segments from every subject, neither of which had any ectopic beat, were analyzed. The statistical differences between the two clinical groups for the first and second segments, and their average were determined for all nine indices. Results showed that there was no significant difference between the two 5-min RR interval segments for any technique. All frequency-domain and non-linear indices, but only one time domain index (SDNN), were significantly different between subject groups. However, some indices were much more sensitive to the clinical differences than others; with the best performing techniques, one non-linear index VLI and one time domain index SDNN, followed by all three frequency indices of LFn, HFn and LF/HF, and finally two of the other non-linear indices VAI and SampEn. A simple RBF SVM-based classification algorithm gave a good performance for classifying the CHF and NSR subjects. And the mean Se, Sp and Acc of SVM classifier from 10 folds were 91.31%, 90.04% and 90.95% respectively. We have shown that there are characteristic differences in heart rate changes between congestive heart failure and normal sinus rhythm, suggesting characteristic rhythm differences. (C) 2018 Elsevier Ltd. All rights reserved.