Deep learning algorithm predicts angiographic coronary artery disease in stable patients using only a standard 12-lead electrocardiogram. (14th October 2021)

• Record Type: Journal Article
• Authors: Leasure, M; Jain, U; Butchy, A; Otten, J; Covalesky, V; McCormick, D; Mintz, G
• Is Part Of: European heart journal. Volume 42(2021)Supplement 1
• Publication Date: 2021-10-14
• Subjects: Noninvasive Diagnostic Methods; Cardiology -- Periodicals; Heart -- Diseases -- Periodicals; 616.12005
• DOI: 10.1093/eurheartj/ehab724.1181
• Languages: English
• ISSNs: 0195-668X

Abstract

Abstract: Background: Current standard electrocardiogram analysis algorithms cannot predict the presence and extent of coronary artery disease (CAD), especially in stable patients. Objectives: This study assessed the ability of a novel artificial intelligence algorithm (ECGio) to predict the presence, location, and severity of coronary artery lesions in an unselected stable patient population. Methods: A cohort of 1659 stable outpatients were randomly divided into training (86%) and validation (14%) subsets, maintaining population characteristics. ECGio was trained and then validated using electrocardiograms paired with angiograms collected from electronic medical records. Coronary artery lesions were classified in two analyses. The primary classification was no/mild (≤30% diameter stenosis [DS]) vs moderate (30–70% DS) vs severe (≥70% DS) CAD; and the secondary classification was yes/no based on ≥50% DS in any vessel. Results: In the primary analysis of the validation cohort, 22 had no angiographic CAD and were grouped with 56 patients with mild CAD (DS ≤30%), 31 had moderate CAD (DS 30–70%), and 113 had severe CAD (DS ≥70%). The weighted average sensitivity was 93.6%, and the weighted average specificity was 96.5%. In the secondary analysis of the validation cohort, 93 had significant CAD; and 128 did not have significant CAD. There was a sensitivity of 93.1% and specificity of 85.6% in determining the presence of clinically significant disease (≥50% DS) in at least one vessel. On a vessel level ECGio was able to predict stenosis severity with average error in the LAD of 18%, the LCX of 19%, the RCA of 18%, and the LM of 8%. Conclusion: This validation study strongly suggests that it is possible to utilize an artificial intelligence algorithm to determine the presence and severity of CAD in stable patients using data from a standard 12-lead electrocardiogram. FUNDunding Acknowledgement: Type of funding sources: None.