Nonlinear dynamic analysis of electrocardiogram on heart rate variability

Abstract

The nonlinear dynamic characteristics of cardiovascular rhythms had been applied to evaluate the cardiac health by some studies. Largest Lyapunov Exponent (LLE) is the general index to depict the signals' nonlinearity and several algorithms were developed for its calculation. LLE also evaluated the nonlinearity of the heart electric activities. In previous studies, several LLE algorithms were utilized to two data formats, original ECG signal and HRV signals, arbitrarily and led to ambiguous and contradictory results. Therefore, in our study, we designed the experiments to explore the applications of two typical LLE algorithms, Wolf and Rosenstein, for ECG and HRV data formats of heart electric signals. The experiments show that Wolf and Rosenstein algorithms are both suitable for ECG signals, but only Wolf can obtain the consistent results on HRV signals. For HRV signals, a pipeline was proposed to analyze the nonlinear dynamic characteristics with LLE, Approximate Entropy and Poincaré plot after denoising and automatic segmenting of steady signals. The pipeline was applied to two HRV datasets from public database and clinical collected data, composed with healthy young and elderly subjects. Experiments achieved consistent results that the heart systems of young people are more complex and less predictable than the elderly. Furthermore, Poincaré plots give clear visual explanations for the increasing nonlinearity among three groups of clinical-healthy hearts of different ages, which offer new sight of the dynamic characteristics for the senile hearts. The pipeline is available at http://ndhrv.codeplex.com.