In Europe and the USA, 50 million patients present to emergency departments with chest pain every year. Regrettably, the current standard of care, which relies on ST-elevation criteria in ECG readings, fails to accurately detect acute coronary occlusion in 25-30% of NSTEMI patients, thereby delaying crucial interventions. On the other hand, hospitals also suffer from 15-20% false positive cath lab activation rates due to false positive ST-elevation criteria. A landmark study published in the European Heart Journal - Digital Health discusses a pioneering AI-powered ECG model designed to enhance the detection of acute coronary occlusion myocardial infarction. 

This model, specifically targeting acute occlusion myocardial infarction (OMI), was developed using a comprehensive dataset of 18,616 ECGs from 10,543 patients suspected of ACS, sourced from an international database with angiographically confirmed outcomes. 

The AI-ECG model was tested against traditional STEMI criteria and expert ECG assessments and showed remarkable efficiency with an AUC of 0.938 for detecting OMI. Achieving a high accuracy of 90.9%, with a sensitivity of 80.6% and a specificity of 93.7%, it significantly outperformed the traditional STEMI criteria and aligned with the proficiency of ECG experts to detect acute coronary occlusion. These published findings are detailed in the European Heart Journal - Digital Health.

The AI model significantly surpassed the current standards in detecting ECG indicators of acute coronary occlusion, warranting cathlab activation. It was found to be twice as sensitive compared to the current standard of care and detected acute coronary occlusion three hours earlier. This suggests its potential to improve chest pain triage, ensuring appropriate and timely referral for immediate revascularization.

"When treating patients with symptoms potentially indicating Acute Coronary Syndrome (ACS), immediate and accurate diagnosis is crucial for timely intervention. The OMI AI Model leverages decades of our ECG morphology research to accurately distinguish acute coronary occlusions from patterns which mimic them, going beyond mere ST-elevation analysis”, states Dr. Stephen W. Smith, faculty physician at the Emergency Medicine Residency, Hennepin County Medical Center and Professor of Emergency Medicine at the University of Minnesota, and author of Dr. Smith’s ECG Blog.

This patient database, featuring a broad spectrum of cases, was instrumental in training the AI model. It provided an abundance of real-world data, enabling the model to recognize subtle and complex patterns associated with acute coronary occlusion myocardial infarction.

The development of this ECG AI model is a product of international collaboration and research involving esteemed institutions, including the world-renowned Cardiovascular Center Aalst in Belgium and the University of Napoli Federico II. 

Dr. Dan Schelfaut, Co-Director and Head of Coronary Care Unit Cardiovascular Center Aalst, OLV Hospital, elaborates on its impact: “This AI innovation symbolizes the fusion of technology with cardiac care, paving the way for swift and accurate OMI diagnosis in diverse clinical scenarios. The AI model's superior accuracy in detecting acute OMI, as compared to the STEMI criteria, underscores its potential to revolutionize ACS triage, ensuring timely and appropriate referrals for immediate revascularization.”

Since its implementation at the Cardiovascular Center Aalst, the AI model has proven its efficacy and reliability in clinical practice. Physicians at the center have observed a notable increase in diagnostic accuracy, substantially improving patient outcomes. This rapid and accurate identification of acute coronary occlusion on the 12-lead ECG is crucial in administering prompt and effective treatment.