Written by Ketan Patel
Spoon Feed
The Queen of Hearts AI (QoH AI) model beat both emergency physicians (EPs) and cardiologists at deciding who needs immediate catheterization for certain STEMI-equivalents and STEMI-mimics—88.9% accuracy vs. ~66%—promising fewer missed occlusion myocardial infarctions (OMIs) and fewer unnecessary activations.
ECG Poker: Can you beat the Queen of Hearts?
The ever-expanding role of AI in ECG interpretation has been examined many times by us here at Journal Feed including a most recent comparison of 3 AI Models. The performance, however, of more advanced AI models specifically trained to analyze ECGs for OMI, even when not obvious STEMI, is unknown when compared to physicians.
This study evaluated how accurate EPs and cardiologists were vs. the ECG-AI for deciding immediate cath lab activation among STEMI-equivalent and STEMI-mimic ECGs. This was a cross-sectional survey of 53 emergency physicians (EPs) and 42 cardiologists interpreting 18 curated ECGs; performance compared with QoH AI against an angiography/troponin/echo/follow-up reference standard. Results showed physician accuracy was 65.6% (EPs; 95%CI 51–78) and 65.5% (cardiologists; 95%CI 51–77) vs AI 88.9% (95%CI 82–93; p<0.001). EP sensitivity/specificity 59%/72%; cardiologists 51%/80%.
Limitations include single-center design, survey-based benchmarking with one example per ECG type and an artificially set ~50% OMI prevalence limiting external validity and risking spectrum bias.
It is also important to note that in the subset of ECGs studied, consensus in management including need, timing and benefits of immediate catheterization remain variable and require further study and validation. The risks of potential “over-activation” are thus hard to quantify.
How does this change my practice?
I’m not ready to throw in the towel on my (or my colleague’s) abilities to discern pathology on ECG. I do, however, understand human limitations, as well current AI limitations and variability among models. I look forward to the day when we can rely on smarter, more accurate, task specific trained AI models – like QoH AI – that will help augment my decision making leading to better resource utilization and patient outcomes.
Source
Accuracy of cath lab activation decisions for STEMI-equivalent and mimic ECGs: Physicians vs. AI (Queen of Hearts by PMcardio). Am J Emerg Med. 2025 Jul 30;97:193-199. doi: 10.1016/j.ajem.2025.07.061. Epub ahead of print. PMID: 40763602.
