Assessment of chaotic parameters in nonstationary electrocardiograms by use of empirical mode decomposition

This study addressed the issue of assessing chaotic parameters from nonstationary electrocardiogram (ECG) signals. The empirical mode decomposition (EMD) was proposed as a method to extract intrinsic mode functions (IMFs) from ECG signals. Chaos analysis methods were then applied to the stationary IMFs without violating the underlying assumption of stationarity. Eight ECG data sets representing normal and various abnormal rhythms were obtained from the American Heart Associate Ventricular Arrhythmia database. The chaotic parameters including Lyapunov exponent, entropy, and correlation dimension were computed. The results consistently showed that the 10th IMF (IMF-10) was stationary and preserved sufficient nonlinearity of the ECG signals. Each IMF-10 from the data sets (n = 8) gave a positive dominate Lyapunov exponent (0.29-0.64, p < 0.0001), a positive entropy (0.039-0.061, p < 0.0001), and a noninteger correlation dimension (1.1-1.9). These were evidences of a chaotic dynamic system. We therefore concluded that the original ECG signals must also have chaotic properties. The chaotic parameters did not show significant differences among the eight data sets representing normal sinus rhythm and various abnormalities. This study has demonstrated an effective way to characterize nonlinearities in nonstationary ECG signals by combining the empirical mode decomposition and the chaos analysis methods.