AI-Enabled ECG Analysis Improves Diagnostic Accuracy and Reduces False STEMI Activations: A Multicenter U.S. Registry

Area of Science:

• Cardiology
• Medical Artificial Intelligence
• Health Informatics

Background:

• Timely reperfusion is crucial for reducing mortality in ST-segment elevation myocardial infarction (STEMI).
• Current ECG-guided cardiac catheterization laboratory (CCL) activation improves response but faces diagnostic uncertainty, leading to false-positive activations (FPAs) and delays, especially with atypical presentations.

Purpose of the Study:

• To evaluate the diagnostic performance of AI-based ECG analysis in real-world STEMI triage.
• To assess the operational impact of AI in multicenter STEMI diagnosis across a U.S. registry.

Main Methods:

• Retrospective analysis of 1,032 patients with suspected STEMI undergoing emergent CCL activation across three U.S. PCI centers (Jan 2020-May 2024).
• Comparison of standard triage with blinded retrospective AI ECG analysis (Queen of Hearts, PMcardio) for detecting acute coronary occlusion and mimics.
• Reference standard: angiographically confirmed culprit lesion with positive enzymes. Analysis included diagnostic accuracy, subgroup performance, and FPA reclassification.

Main Results:

• AI ECG analysis demonstrated superior sensitivity (92.0%) versus standard triage (71.0%) for STEMI detection (p < 0.001).
• AI significantly reduced false-positive activation rates (7.9% vs. 41.8%) and improved specificity (81.0% vs. 29.0%) (p < 0.001).
• AI achieved an AUC of 0.94, maintaining performance across challenging subgroups and correctly reclassifying 91% of biomarker-negative FPAs.

Conclusions:

• AI-based ECG analysis substantially enhances STEMI detection accuracy and reduces unnecessary interventions.
• The AI model effectively identifies non-conventional STEMI presentations, addressing diagnostic uncertainty.
• Findings support the integration of AI-ECG analysis into acute chest pain management pathways for improved patient outcomes.