The Cardiodynamicsgram Based Early Detection of Myocardial Ischemia Using the Lempel-Ziv Complexity

Background: Electrocardiogram (ECG) is a routine method for detecting myocardial ischemia in clinical practice, but more than half of ECGs are without specific ischemic changes. Cardiodynamicsgram (CDG) is an effective method to detect ischemia with non-diagnostic ECG. The Lyapunov exponent (LYE) and the Fourier transform coefficient are combined to characterize the spatial and temporal features of CDG. However, in some cases, the Lyapunov exponent does not accurately enough describe the degree of irregular morphology of CDG for ischemic patients. In this context, this study aims to improve the characterization of CDG using the Lempel-Ziv (LZ) complexity instead of the Lyapunov exponent. Methods: The cardiodynamics information inside ECG is extracted via deterministic learning from the ST-T segments of ECG and then the CDG is generated by plotting the extracted three-dimensional cardiodynamics information. The Lyapunov exponent and LZ complexity are calculated from CDG and coupled with the Fourier transform coefficient respectively to construct the LYE model and LZ model for detecting myocardial ischemia. Results: 393 subjects presenting non-diagnostic ECG are enrolled in the study. 312 of them are ischemic patients selected as the myocardial ischemia group, and the other 81 non-ischemic subjects are selected as the healthy control group. The average sensitivity, specificity, and accuracy of the LYE model and the LZ model are 90.7% vs 93.4%, 86.4% vs 86.8%, and 89.0% vs 90.8%, respectively. Meanwhile, the proposed method achieves better performance on the PTB database than most of the previous studies in detecting ischemia or infarction. Conclusion: The results indicate that LZ complexity can accurately characterize the cases that cannot be accurately depicted by Lyapunov exponent, and the corresponding model is more accurate for the early detection of myocardial ischemia.