Multiscale Joint Recurrence Quantification Analysis Integrating ECG Spatiotemporal and Dynamic Information for Cardiopathy Detection

Cardiovascular disease disturbs the structure and function of myocardial cells, altering the spatiotemporal and dynamic patterns of cardiac electrical activity. The conventional joint recurrence analysis contributes to revealing the spatiotemporal pattern of heart disease-induced changes, but it usually overlooks the intrinsic dynamics of the heart. In clinical practice, multiple non-cardiovascular factors can cause similar spatiotemporal alterations in the heart's electrical activity that those observed in cardiovascular disease. Therefore, accurately diagnosing cardiovascular disease remains a challenging clinical problem. In this paper, we propose a novel multi-scale joint recurrence quantification analysis (MSJRQA) approach that integrates spatiotemporal and dynamic information of ECG signals to distinguish the variations in ECG caused by diverse factors. The ECG signal is first modeled using deterministic learning to capture dynamic information. ECG signal and its dynamics are decomposed into multiple scales by variational mode decomposition (VMD). Subsequently, the spatiotemporal and dynamic information is utilized as input for joint recurrence quantification analysis (JRQA) to characterize spatiotemporal and dynamical patterns of disease induction in ECG signals. Finally, an ensemble classifier is applied to differentiate myocardial infarctions (MI) from healthy individuals and non-MI patients accompanied by ST-T changes. The experimental results demonstrate that the proposed MSJRQA method yields superior results with 94.50% and 90.78% accuracy for the identification of MI and the differentiation of diseases, respectively. Therefore, it can effectively assist cardiologists in the early diagnosis of MI and decision-making.