Abstract Body (Do not enter title and authors here): INTRODUCTION

Current ACS guidelines suggest classifying patients according to the presence of persistent ST segment elevation, as a finding suggestive of acute thrombotic coronary occlusion. However, large series have documented that up to 15% of patients initially classified as NSTEMI will show evidence of total coronary occlusion on index angiography, increasing the length of stay, use of hospital resources, and short-term and long-term mortality. Therefore, prompt detection of Acute Coronary Occlusion Myocardial Infarction (ACOMI) is paramount.

AIMS

We aimed to assess the performance of an AI-ECG based model capable of detecting ACOMI in the setting of patients with ACS.

METHODS

This is a prospective study based on the development of an AI-ECG based model capable of detecting ACOMI. A publicly available dataset (PTB-XL ECG) of 21,837 12-lead ECGs was used for training in recognizing ST-segment elevation. Regarding the detection of ACOMI, 12-lead ECGs from 361 patients who presented to the ED with an ACS (2017-2023) at our center were digitized with phone cameras of varying quality. ECGs were independently evaluated by two expert cardiologists blinded to clinical outcomes; each was asked to determine a) whether the patient had an STEMI, based on universal criteria or b) if STEMI criteria was not met, to identify any other ECG finding suggestive of ACOMI. ACOMI was defined as the presence of one of the following: TIMI V thrombus, TIMI thrombus grade 2 or higher + TIMI grade flow 1 or less, or the presence of a subocclusive (>90%) lesion. Patients were classified into four groups: STEMI + ACOMI, NSTEMI + ACOMI, STEMI + non-ACOMI and NSTEMI + non-ACOMI. Performance of the AI model was evaluated using a comparison of multiple areas under the receiver operating characteristic curve (AUC-ROC). Sensitivity, specificity, positive and negative (PPV, NPV) predictive values and F1-score were also calculated.

RESULTS

The AI model accomplished an AUC of 0.8667 in identifying ACOMI, outperforming ECG experts (AUC: 0.3333) and the use of universal STEMI criteria (AUC: 0.5095). It also accomplished a sensitivity of 1, specificity of 0.733, a PPV of 0.846, an NPV of 1 and an F1-score of 0.92.

CONCLUSION

Our AI-ECG model demonstrated a higher diagnostic precision for the detection of ACOMI compared with experts and use of STEMI criteria. Further research and external validation is needed to understand the role of AI-based models in the setting of ACS.