Abstract 4364391: Artificial Intelligence-Powered Electrocardiogram to Detect Acute Coronary Obstruction in Patients with Suspected Acute Coronary Syndrome

Abstract

Background: ST-elevation on a 12-lead ECG helps identify occlusion myocardial infarction (OMI) requiring emergent invasive coronary angiography (ICA) and revascularization. However, ~25% of non-ST-elevation myocardial infarctions have OMI on ICA, highlighting the need for improved risk assessment. Methods: This multicenter, retrospective study analyzed consecutive adults with suspected high-risk non-ST-elevation acute coronary syndrome (NSTE-ACS) from 2022 to 2024, using the Minneapolis Heart Institute Level 2 Protocol within the PROGRESS-CATH registry. Eligible patients underwent a 12-lead ECG before urgent ICA referral based on criteria such as ST depression, anterior T-wave inversion, positive biomarkers, unstable arrhythmias, clinical instability, or ischemia on stress testing. STEMI (Level 1) patients were excluded. The FDA-cleared OMI AI ECG model (Powerful Medical, Slovakia) assessed OMI risk, which was validated via ICA. OMI was defined as an angiographic culprit lesion with TIMI-0-2 flow or TIMI-3/unknown flow with elevated cardiac troponin (cTn) or new regional wall motion abnormality. Results: Among 172 patients with suspected NSTE-ACS, 76 (44%) were classified as high-risk of OMI by the AI model, while 96 (56%) were categorized as non-OMI. Among the 76 patients with high-risk of OMI by AI, 65 (86%) had OMI confirmed by the ICA, whereas 11 (15%) did not have a culprit lesion. Patients in the AI-based OMI and non-OMI groups had comparable demographic characteristics. Institutional treatment protocols ensured similar management across both groups. Time intervals between admission and first ECG (21 [10-29] vs. 14 [7.3-60] minutes; p=0.93), first cTn measurement (42 [32-57] vs. 56 [26-98] minutes; p=0.78), and presentation to the catheterization laboratory (194 [64-850] vs. 250 [68-778] minutes; p=0.80) were comparable. In-hospital mortality was higher in the OMI group (8 patients, 10.5%) compared with the non-OMI group (2 patients, 2.1%) (OR 5.52; 95% CI 1.14-26.82; p=0.03). After a mean follow-up period of 442 days, there was higher mortality in the OMI group (HR 3.18; 95% CI 1.05-10.1; p=0.04). Conclusion: AI-enhanced detection of OMI among patients without ST-elevation facilitates the rapid detection of patients at high risk for mortality that may require prompt revascularization and may outperform standard 12-lead ECG risk-stratification.