Detection of Coronary Artery Disease Based on Amplitude and Interval Analysis of ECG and PCG Signals

Coronary Artery Disease (CAD) is a significant cause of increased mortality worldwide. Long-term monitoring and accurately assessing can effectively reduce the mortality rate associated with CAD. Electrocardiogram (ECG) and phonocardiogram (PCG) are simple, affordable, and non-invasive approaches that reflect the heart's physiological state, with ECG detecting electrical activity abnormalities caused by ischemia and PCG providing insights into mechanical function and contractility. This study simultaneously recorded ECG and PCG signals from 214 subjects, who were categorized into four groups based on varying degrees of coronary artery stenosis: severe CAD (sCAD, >70%), mild CAD (mCAD, 30%-70%), suspected CAD (susCAD, <30%), and healthy control (Health). To identify pathological changes in ECG and PCG signals, our approach locates key points in ECG (P, Q, R, S, T) and PCG (S1, S2, systole, diastole), and extracted various interval sequences (RR, QT, PR, S1, S2, systole, diastole) and amplitude sequences (P, Q, R, T, ST, S1, S2, systole, diastole). Subsequently, multi-domain features, including time-domain, frequency-domain, and nonlinear features, were computed, after which Recursive Feature Elimination (RFE) was employed to select the optimal feature subset, while Support Vector Machine (SVM) was utilized to identify CAD severity. Result showed that the joint analysis outperforms traditional single-signal methods in detecting CAD severity, and the addition of amplitude features can extract more pathological information and enhance classification performance, achieving accuracies of 93.82%, 97.69%, and 97.14% for the three degrees of coronary artery disease, respectively. Furthermore, this study confirms the feasibility of using ECG and PCG signals for the early screening of coronary artery disease.