Use of Electrocardiograms to Identify Coronary Artery Disease

BACKGROUND: The current gold standard for the diagnosis of coronary artery disease (CAD) is invasive angiography; however, it is an invasive procedure. Therefore, we developed an artificial intelligence model designed to predict significant CAD from a resting digital 12-lead electrocardiogram (ECG). OBJECTIVES: This retrospective study assessed the model's ability to predict clinically significant CAD in a patient population presenting for coronary angiography. METHODS: From 2019 to 2021, 16,476 patients had a resting 12-lead digital ECG recorded within 90 days prior to coronary angiography. The artificial intelligence model was developed using 10-fold cross-validation methodology. Clinically significant disease was defined as angiographic diameter stenosis ≥70% in the left anterior descending, left circumflex, or right coronary artery or ≥50% in the left main coronary artery. We then applied the model to an external validation set. RESULTS: In the cross-validation cohort, the prevalence of clinically significant CAD was 64.5%; the model achieved a positive predictive value of 91.7% (95% CI: 89.9%-93.4%), negative predictive value of 72.8% (95% CI: 69.6%-76.0%), and area under the curve of 91.4% (95% CI: 89.4%-94.4%) in predicting clinically significant CAD. In external validation, the prevalence of clinically significant CAD was 36.0%; the model achieved a positive predictive value of 82.5% (95% CI: 75.9%-89.2%), negative predictive value of 88.1% (95% CI: 84.0%-92.1%), and area under the curve of 92.4% (95% CI: 89.7%-95.1%) in predicting clinically significant CAD. CONCLUSIONS: This study demonstrated the clinical utility of a deep learning artificial intelligence algorithm to analyze a digital 12-lead ECG to predict the presence of clinically significant CAD as determined by coronary angiography.