AI-Enhanced Electrocardiogram for Detection of Occlusive Myocardial Infarction in High-Risk Non-ST-Segment Elevation...

Whether an artificial intelligence-enhanced electrocardiogram (AI-ECG) improves detection of occlusive myocardial infarction (OMI) in non-ST-segment elevation acute coronary syndrome (NSTE-ACS) compared with standard of care (SOC) remains uncertain.

The objective of the study was to evaluate an AI-ECG model for detecting OMI among patients with suspected high-risk NSTE-ACS using invasive coronary angiography (ICA) as the reference.

Consecutive adults undergoing emergent ICA under the institutional level 2 protocol for suspected high-risk NSTE-ACS (2022-2024) were included. Patients with ST-elevation myocardial infarction were excluded. Initial 12-lead ECGs were retrospectively analyzed using an investigational AI-ECG model (Queen of Hearts, Powerful Medical) validated against ICA findings. OMI was defined as an angiographic culprit lesion with TIMI flow grade 0 to 2 or TIMI flow grade 3 with marked troponin elevation or new regional wall motion abnormality.

Among 224 patients, OMI was present in 129 (58%). AI-ECG identified 75 (58%) of these as OMI and classified 96 patients (43%) as OMI overall. For rule in, specificity was 78% (95% CI: 68%-86%) and positive predictive value 78% (95% CI: 69%-86%). For rule out, sensitivity was 58% (95% CI: 49%-66%) and negative predictive value 57% (95% CI: 48%-66%). Serial ECGs reduced false negatives from 42% to 34%. Compared with the standard of care, an AI-augmented triage approach reduced false positives from 42% to 22% (P < 0.001).

In high-risk NSTE-ACS, AI-ECG improved rule in accuracy for OMI, potentially enhancing early identification of patients requiring urgent ICA. However, rule out performance from initial ECG alone was limited, supporting the need for serial ECGs and clinical judgment.