Using Machine Learning for Prediction of Obstructive CAD

Study Questions:

Does machine learning utilizing clinical factors and coronary artery calcium score (CACS) improve prediction of obstructive coronary artery disease (CAD) coronary computed tomography angiography (CCTA)?


The CONFIRM (COronary CT Angiography EvaluatioN For Clinical Outcomes: An InteRnational Multicenter) registry is a prospective, observational registry that enrolled patients in 12 medical centers across six countries (Canada, Germany, Italy, Korea, Switzerland, and the United States). Data collected included demographic, clinical, and imaging parameters for patients who underwent ≥64-detector row CCTA evaluation because of either suspected or previously established CAD. Of the total cohort, 13,054 patients (≥30 years of age) were identified, for whom clinical data and the CACS were available. Patients with prior CAD or revascularization (percutaneous or surgical) were excluded. A boosted ensemble algorithm (XGBoost) was used, with data split into a training set (80%) on which 10-fold cross-validation was done and a test set (20%). Performance was assessed of the: 1) machine learning model (using 25 clinical and demographic features), 2) machine learning + CACS, 3) CAD consortium clinical score, 4) CAD consortium clinical score + CACS, and 5) updated Diamond-Forrester score.


Of the total 35,281 participants in the cohort, 13,054 patients (ages ≥30 years) had completed CCTA imaging for either suspected or previously established CAD. Of the 13,054 patients, 2,380 (18.2%) had obstructive CAD (≥50% stenosis). Machine learning with CACS produced the best performance (area under the curve [AUC] of 0.881) compared with machine learning alone (AUC of 0.773), CAD consortium clinical score (AUC of 0.734), and with CACS (AUC of 0.866) and updated Diamond-Forrester score (AUC of 0.682); p < 0.05 for all comparisons. CACS, age, and gender were the highest-ranking features.


The investigators concluded that a machine learning model incorporating clinical features in addition to CACS can accurately estimate the pretest likelihood of obstructive CAD on CCTA. In clinical practice, the utilization of such an approach could improve risk stratification and help guide downstream management.


This study highlights the potential for the clinical utility of machine learning to identify patients at higher risk for CAD events. However, as the authors note, such techniques can be limited by the quality of data utilized. Furthermore, the present data are limited to patients who were referred for CCTA; thus, selection bias is likely present. Understanding how best to incorporate such techniques into clinical practice will require rigorous studies from multidisciplinary teams of investigators.

Clinical Topics: Invasive Cardiovascular Angiography and Intervention, Noninvasive Imaging, Prevention, Atherosclerotic Disease (CAD/PAD), Interventions and Coronary Artery Disease, Interventions and Imaging, Angiography, Computed Tomography, Nuclear Imaging

Keywords: Area Under Curve, Constriction, Pathologic, Coronary Angiography, Coronary Artery Disease, Diagnostic Imaging, Learning, Plaque, Atherosclerotic, Primary Prevention, Risk Assessment, Tomography, Emission-Computed

< Back to Listings