Accuracy of cath lab activation decisions for STEMI-equivalent and mimic ECGs: Physicians vs. AI (Queen of Hearts by PMcardio)
Steven Shroyer, Sumeru Mehta, Nandish Thukral, Kyle Smiley, Nathaniel D. Mercaldo, H. Pendell Meyers, Stephen W. Smith
Abstract
INTRODUCTION: Accurate ECG interpretation is crucial to identify occlusive myocardial infarction (OMI) to determine the need for immediate catheterization laboratory activation (CLA). STEMI-equivalent and STEMI-mimic ECG patterns deviate from conventional STEMI criteria, risking misclassification of OMI cases. The diagnostic accuracy for these complex ECGs is unknown. OBJECTIVES: This study aimed to measure physician accuracy for interpreting STEMI-equivalent and STEMI-mimic ECGs for catheterization laboratory activation (CLA) and compare their performance to a machine learning-based artificial intelligence algorithm, Queen of Hearts AI (QoH AI). METHODS: Fifty-three EPs and 42 cardiologists interpreted 18 ECGs (eight STEMI-equivalents, eight STEMI-mimics, with one STEMI, and a normal ECG as controls) to determine the presence of OMI requiring immediate CLA. The same ECGs were analyzed by QoH AI. Interpretations were compared against a reference standard based on angiography, troponin, echocardiography, and clinical follow-up. RESULTS: Interpretation accuracies were similar between EPs and cardiologists (65.6 %, 95 % CI [51, 78]; 65.5 %, 95 % CI [51, 77], respectively; p = 0.969), and significantly lower than QoH AI (88.9 %, 95 % CI [82, 93]) vs. physicians overall, 65.6 %, 95 % CI [52, 77]; p < 0.001). Physicians most frequently misclassified de Winter, Transient STEMI, Hyperacute T-wave OMI, and bundle branch block ECGs. QoH AI only misclassified left bundle branch block with OMI and left ventricular aneurysm without OMI. CONCLUSION: Physicians frequently misinterpret STEMI-equivalent and STEMI-mimic ECGs, potentially impacting CLA decisions. QoH AI demonstrated superior accuracy, suggesting a potential to reduce missed OMIs and unnecessary catheterization laboratory activations. Prospective studies are needed to validate these findings in clinical practice.