Prediction of Coronary Artery Disease Using Machine Learning

Abstract

Coronary artery disease (CAD) is the most common cardiovascular disease. It involves the reduction of blood flow to the myocardium due to the build-up of atherosclerosis in the coronary arteries. We developed 4 machine learning models (namely, logistic regression, random forest, support vector classifier, and deep neural network) to predict the diagnosis of CAD. From January 2017 to December 2018, a total of 372 (265 men and 107 women) eligible patients were enrolled. Two hundred and thirty-two patients (62.4%) were finally confirmed to have CAD via cardiac catheterization examination. Patients' clinical information and semiquantitative parameters from the myocardial perfusion image (MPI) were used as predictors. The accuracy scores were 0.7823 (+/- 0.0496) with logistic regression, 0.7633 (+/-0.0930) with random forest, 0.7718 (+/- 0.0810) with support vector classifier, and 0.7297 (+/- 0.0494) with deep neural network.