Classification of Coronary Artery Disease Using Radial Artery Pulse Wave Analysis via Machine Learning

Abstract

Background Coronary artery disease (CAD) is a major global cardiovascular health threat and the leading cause of death in many countries. The disease has a significant impact in China, where it has become the leading cause of death. There is an urgent need to develop non-invasive, rapid, cost-effective, and reliable techniques for the early detection of CAD using machine learning (ML). Methods 608 participants were divided into three groups: healthy, hypertensive, and CAD. The raw data of pulse wave from those participants was collected. The data were de-noised, normalized, and analyzed using several applications. Seven ML classifiers were used to model the processed data, including Decision Tree (DT), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), Extra Trees (ET), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting (LightGBM), and Unbiased Boosting with Categorical Features (CatBoost). Results The Extra Trees classifier demonstrated the best classification performance. After tunning, the average results were 86.6% accuracy, 91.36% AUC, 86.6% recall, 87.27% precision, 86.58% F1 score, 0.7984 kappa coefficient, and 0.8018 MCC. The macro-average AUC of validation result for independent test set is 94%. The top 10 feature importances of ET model are w/t₁, t₃/t_max, t_max, t₃/t₁, As, h_f/3, t_f/3/t_max, t_f/5, w and t_f/3/t₁. Conclusion Radial artery pulse wave can be used to identify healthy, hypertensive and CAD participants by using Extra Trees Classifier. This method provides a potential pathway to recognize CAD patients by using a simple, non-invasive, and cost-effective technique.