Reliance on STEMI criteria alone misses a substantial proportion of LAD occlusions, while expert and AI ECG interpretation achieved earlier and more accurate detection, with 100% sensitivity for LAD occlusion MI in this study.
Oh, my! AI caught an OMI!
This study evaluated the diagnostic accuracy of guideline-based STEMI ECG criteria vs. blinded expert interpretation and an AI algorithm for identifying total LAD occlusion. This was a sub-analysis of the DOMI-ARIGATO retrospective case-control database. The study included patients presenting with ACS symptoms to two academic EDs, with OMI confirmed angiographically. Among 53 patients with total LAD occlusion, 20 (38%) never met STEMI criteria on any ECG (including serial ECGs). Those true OMIs not meeting STEMI criteria had significantly longer door-to-balloon times (97 vs. 40 minutes, P < 0.001) despite similar infarct sizes and ejection fractions compared with those meeting STEMI criteria. Expert ECG interpretation and an AI-powered ECG model BOTH individually demonstrated 100% sensitivity on the first ECG for detecting LAD occlusion, whereas guideline-based STEMI criteria identified only 62%. Subtle OMI features included hyperacute T waves (85%), pathologic Q waves (70%), and reciprocal changes (50%). Findings underscore the limitations of STE-based diagnosis, support the OMI/NOMI framework, and highlight potential for AI integration to reduce treatment delays in anterior occlusion MI.
How this will change my practice?
This article addresses a central question—how can we better diagnose occlusion MI in patients without significant ST-elevation? This study highlights the need for improved ECG diagnostics. An AI tool designed to detect OMI outside of STEMI criteria achieved 100% sensitivity—twice that of guideline STEMI criteria! While AI is not yet standard of care, this suggests it could eventually support faster OMI recognition, and it reinforces the OMI/NOMI paradigm. Our goal is not only to “rule in STEMI,” but to identify any acute coronary occlusion as early as possible. My interest is piqued… this adds to the growing body of work exploring how AI might be incorporated into emergency medicine to improve outcomes.
Source
Failure of standard contemporary ST-elevation myocardial infarction electrocardiogram criteria to reliably identify acute occlusion of the left anterior descending coronary artery. Eur Heart J Acute Cardiovasc Care. 2025 Aug 6;14(7):403-411. doi: 10.1093/ehjacc/zuaf037. PMID: 40717627.
