A deep learning ECG model identified occlusion myocardial infarction with a C-statistic of ≥0.95 and non-OMI infarctions with a C-statistic of ≥0.87, while also localizing culprit lesions.
Observational (n=540,372)
Does a deep learning ECG model accurately identify and localize occlusion myocardial infarction?
A deep learning model applied to emergency ECGs can accurately identify and localize occlusion myocardial infarction, potentially expediting reperfusion therapy.
Effect estimate: C-statistic ≥0.95 for OMI, ≥0.87 for non-OMI
Abstract Rapid identification and localization of an acute coronary occlusion are vital to prevent myocardial damage, yet reliance on ST-segment ECG criteria misses many acute occlusion myocardial infarctions (OMI) and triggers unnecessary acute angiographies. Here, we present a trained and validated deep learning model using 540,372 emergency ECGs paired with definitive catheterization outcomes. The model has a C-statistic of ≥0.95 for OMI and ≥0.87 for non-OMI infarctions and can localize culprit lesions in the three main coronary branches, which can guide the angiographer. Performance is similar across age, sex, and ECG hardware subgroups. Obviating dependence on ST-elevations and troponins, this model for the identification and localization of OMI has the potential to shorten the time to reperfusion of an acute coronary occlusion and save resources. Because human oversight of OMI detection on the ECG is limited, randomized clinical trials with patient-relevant outcomes are warranted.
Published May 13 2026; high social shares among cardiologists on X and ResearchGate; discussed in AI cardiology threads.
Gustafsson et al. (2026) conducted an observational in Occlusion myocardial infarction (n=540,372). Deep learning ECG model was evaluated on Identification of occlusion myocardial infarction (OMI) and non-OMI infarctions (C-statistic ≥0.95 for OMI, ≥0.87 for non-OMI). A deep learning ECG model identified occlusion myocardial infarction with a C-statistic of ≥0.95 and non-OMI infarctions with a C-statistic of ≥0.87, while also localizing culprit lesions.