AI-enhanced ECG for acute coronary syndrome triage: A state-of-the-art review

Acute coronary syndrome (ACS) remains a leading cause of emergency department presentations, yet triage based on the ST-elevation myocardial infarction (STEMI)/non-ST-elevation myocardial infarction (NSTEMI) paradigm misses approximately 25-34% of acute coronary occlusion myocardial infarction (OMI). Early artificial intelligence-electrocardiography (AI-ECG) models showed retrospective promise for detecting ischemic ECG patterns but lacked prospective validation. This review synthesizes emerging multicenter registry, prospective cohort, and pathway trial evidence for AI-ECG in ACS triage, including the Queen of Hearts registry, ROMIAE, and DIFOCCULT-3 studies. It is essential to distinguish two separate clinical tasks: (1) detecting OMI for emergent catheterization laboratory activation, and (2) ruling out acute myocardial infarction (MI), which requires serial high-sensitivity troponin and cannot be achieved by ECG alone. Contemporary findings suggest AI-ECG significantly improves OMI detection sensitivity (92% vs. 71% for standard care) and reduces false-positive catheterization laboratory activations by up to 91% among biomarker-negative patients. For acute MI rule-out, AI-ECG shows promise as an adjunct to troponin-based strategies, with a negative predictive value of approximately 99% when combined with high-sensitivity troponin and clinical risk scores. We propose a 'Second Opinion' framework in which AI augments physician judgment as a clinical decision support tool. Key implementation challenges include algorithmic bias, alert fatigue, documentation, and the risk of widening the digital divide. AI-ECG represents a shift toward a physiologically driven OMI vs. non-occlusive myocardial infarction (NOMI) diagnostic framework.