Skip to content
Review

Abstract 9731: Artificial Intelligence-Enabled Electrocardiography for Cardiovascular Diagnosis and Risk Prediction: A Systematic Review

Nov 2026 · Circulation · 0 citations

Abstract

Background: Artificial intelligence (AI)-enabled electrocardiography (ECG) may extend a low-cost, widely available test beyond conventional interpretation, but the evidence base remains heterogeneous and variably validated. Methods: We conducted a PRISMA 2020-compliant systematic review of original human studies evaluating AI, machine learning, or deep learning for ECG-based cardiovascular detection, classification, monitoring, or risk prediction. PubMed and the Cochrane Library were searched from inception through January 2024. Risk of bias was assessed using PROBAST. Because of heterogeneity in populations, ECG modalities, model architectures, validation strategies, and outcomes, findings were synthesized narratively. Results: One hundred studies were included. Arrhythmia detection, particularly atrial fibrillation, was the most frequently studied and most mature application. Large externally validated multilabel ECG-classification studies reported AUROC values of 0.97-0.99, while AI-ECG models for reduced left ventricular ejection fraction screening achieved AUROC values of 0.918-0.93. Structural phenotype recognition showed AUROC values of 0.77-0.94 depending on the target condition. Benchmark-based arrhythmia studies often reported accuracy or AUROC values above 0.95, but clinically representative cohorts generally produced more moderate estimates. Conclusions: AI-enabled ECG analysis has expanded from rhythm classification to structural disease screening and cardiovascular risk prediction. Arrhythmia detection and automated ECG classification appear closest to clinical deployment. Prospective multicenter studies with external validation, calibration assessment, and workflow-integrated evaluation are needed before broader adoption.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.