Key takeaways
- Accurate diagnosis of an occlusive myocardial infarction (MI) can be challenging in patients who do not have a characteristic ECG sign called an ST elevation.
- This study assessed whether an artificial intelligence (AI) tool could help detect occlusive MI using the first ECG of patients without an ST elevation.
- AI-based interpretation outperformed standard approaches, suggesting that it may complement existing tools to improve accurate, early diagnosis and timely treatment.
Lisbon, Portugal – 20 March 2026: Artificial intelligence (AI)-based ECG interpretation outperformed standard pathways for the detection of occlusive myocardial infarction (MI), according to a study presented today at ESC Acute CardioVascular Care 2026,[1] the annual congress of the Association for Acute CardioVascular Care (ACVC), a branch of the European Society of Cardiology (ESC).
In patients with suspected acute coronary syndrome (ACS), a specific change on an ECG, called an ST elevation, is an indicator that the patient may have an occlusion in coronary artery. This type of heart attack is known as an ST-elevation myocardial infarction (STEMI) and it requires immediate percutaneous coronary intervention to restore the heart’s blood flow. In patients who do not have an ST elevation, the cause of the chest pain can be less certain and further tests are needed to confirm if the MI is due to an occlusion.
Presenter, Doctor Federico Nani from Central Hospital Bolzano, Italy, explained: “Many patients without an ST elevation have an occlusive MI, but it can be difficult for clinicians to quickly and accurately recognise this, leading to delays in providing emergency treatment. We investigated whether AI-based interpretation of the initial ECG could improve the accuracy of detecting occlusive MIs in the absence of an ST elevation to optimise patient management.”
This single-centre prospective study included 1,490 patients who had symptoms suggestive of ACS but without an ST elevation on the initial ECG. The mean age was 63 years and 42% were female. Clinicians interpreted the initial ECG, tested levels of the cardiac biomarker troponin, and performed coronary angiography, when required, to diagnose occlusive MI based on ESC Guidelines. In parallel, the initial ECG was interpreted by a smartphone-based CE-certified AI-ECG algorithm.
AI-based ECG interpretation ruled out occlusive MI in 1,382 patients and detected it in 108 patients (7%). The AI-based method correctly identified obstructive MI in 84% of cases. Sensitivity was 77%, specificity was 99% and the negative predictive value was 98%. There were 27 false negatives (2%) and 17 false positives (1%).
According to the standard diagnostic pathway, occlusive MI was ruled out in 1,207 patients based on troponin levels and 283 patients underwent coronary angiography to confirm or exclude the diagnosis. Overall, human ECG-interpretation correctly identified occlusive MI in 42% of cases.
Doctor Nani concluded: “This simple, accessible AI-based approach demonstrated superior accuracy in identifying and excluding occlusive MI compared with conventional diagnostic pathways in patients without an ST elevation. The results of our single-centre study require further validation, but these findings suggest that AI ECG interpretation is a valuable addition to existing decision-making tools to improve early recognition and timely, effective treatment.”
The power of AI to support the management of cardiovascular disease will be further explored as the spotlight theme of this year’s annual ESC Congress, taking place from 28–31 August 2026 in Munich, Germany.
ENDS
