In a large, multi-center evaluation presented as Late-Breaking Science at the TCT 2025 conference, investigators assessed the diagnostic accuracy of the Queen of Hearts™ AI algorithm for ST-segment elevation myocardial infarction (STEMI) detection in emergency care. The study compared AI-enhanced ECG interpretation against standard triage protocols across three U.S. PCI centers, encompassing more than 1,000 patients who activated emergent reperfusion pathways. Published in JACC: Cardiovascular Interventions, the results demonstrated significantly improved accuracy and reduced false activations when using AI-driven analysis.
Published in: JACC Cardiovascular Interventions
Published on: 28 October 2025
Background
Timely and accurate recognition of STEMI remains critical for reperfusion success and patient survival. Traditional ECG-based triage systems, though foundational, often misclassify cases due to variability in ECG morphology and interpretation—particularly in atypical presentations. This diagnostic uncertainty leads to both delayed reperfusion and excessive false activations of catheterization laboratories, straining healthcare systems.
The Queen of Hearts™ algorithm, developed by Powerful Medical, applies deep learning to ECG interpretation, offering explainable AI insights into ischemic changes. Previous data from the ongoing DIFOCCULT-3 Randomized Controlled Trial (RCT) suggested that AI support could shorten reperfusion times by up to five hours. The present study aimed to validate these findings in real-world U.S. clinical environments.
Methods
Investigators analyzed 1,032 consecutive patients who triggered emergent reperfusion protocols across three geographically distinct tertiary PCI centers—Beth Israel Deaconess Medical Center (Boston), UC Davis Medical Center (Sacramento), and UTHealth Houston. Each ECG underwent parallel interpretation using standard triage criteria and the Queen of Hearts™ algorithm.
Primary outcomes included the proportion of true STEMIs correctly identified on the initial ECG and the false activation rate for non-STEMI cases. Secondary analyses examined cross-institutional consistency, demographic variability, and operational impact. Data from the DIFOCCULT-3 RCT were referenced for comparison, encompassing 6,000 patients with acute coronary syndromes across 18 PCI hospitals in Turkey.
Results
The Queen of Hearts™ AI model identified 92% of true STEMIs on first ECG assessment versus 71% under standard triage (Δ +21%, p<0.001). False-positive cath-lab activations declined from 42% to 8%—a fivefold reduction. The algorithm demonstrated consistent performance across all sites, patient subgroups, and workflow settings.
Secondary data from DIFOCCULT-3 showed that AI-guided triage accelerated reperfusion by up to five hours and improved short-term clinical outcomes, confirming real-world applicability. Long-term survival analyses are ongoing, with results expected in 2026.
Conclusion
AI-enhanced ECG interpretation using the Queen of Hearts™ algorithm significantly improves STEMI detection accuracy while minimizing unnecessary activations. These results validate the model’s potential to expedite diagnosis, optimize transfer workflows from non-PCI centers, and improve patient outcomes. By combining interpretive precision with explainable visualization, the technology represents a paradigm shift in acute cardiac care—bridging diagnostic gaps and ensuring timely, lifesaving interventions.
