Mass exponent spectrum analysis of human ECG signals and its application to complexity detection

The complexity of electrocardiogram (ECG) signal may reflect the physiological function and healthy status of the heart. In this paper, we introduced two novel intermediate parameters of multifractality, the mass exponent spectrum curvature and area, to characterize the nonlinear complexity of ECG signal. These indicators express the nonlinear superposition of the discrepancies of singularity strengths from all the adjacent points of the spectrum curve and thus overall subsets of original fractal structure. The evaluation of binomial multifractal sets validated these two variables were entirely effective in exploring the complexity of this time series. We then studied the ECG mass exponent spectra taken from the cohorts of healthy, ischemia and myocardial infarction (MI) sufferer based on a large sets of 12 leads' recordings, and took the statistical averages among each crowd. Experimental results suggest the two values from healthy ECG are apparently larger than those from the heart diseased. While the values from ECG of MI sufferer are much smaller than those from the other two groups. As for the ischemia sufferer, they are almost of moderate magnitude. Afterward, we compared these new indicators with the nonlinear parameters of singularity spectrum. The classification indexes and results of total separating ratios (TSR, defined in the paper) both indicated that our method could achieve a better effect. These conclusions may be of some values in early diagnoses and clinical applications.