Artificial Intelligence Based Timing, Infarct Size and Outcomes in Acute Coronary Occlusion Myocardial Infarction

Overview

The present study is practice-driven and merely observational and prospective. In clinical routine, patients who suffer from suspected ACS and do not show ST elevation in the ECG, different timing proposals in the guidelines and logistically driven differences lead to considerably variable timings in invasive coronary anatomy assessments. This handling may lead to larger infarct sizes when OMI is overseen. Therefore, the present study aims to observe a) whether an AI model is capable of correctly identify OMI in eligible patients and b) if in these patients troponin peak levels vary depending on the elapsed time between OMI diagnosis and coronary intervention. As the model has not been established yet clinically and in the guidelines, it is safe to assume the usual pathway from first medical contact to specialist's attention is undertaken. When a patient presents in an emergency department or places an emergency call, the physicians assess the situation as usal and as stated in the current guidelines1. If no STEMI is confirmed, the NSTE-ACS protocol is started. The patients who are ruled out for ACS are excluded from the final analysis (screening). In this case, the AI model is tested on their ECG in order to assess whether there are false positives. The patients which are in the ACS "rule-in" trail and undergo final coronary angiography will naturally be divided in patients which were classified as OMI and as non-OMI by the AI model. Furthermore, they will present a different "Time from OMI diagnosis to PCI) and variable troponin peak levels. By leveraging this natural variability, a practical distinction and multiple analyses can be done: 1. The feasibility of AI-powered ECG interpretation in the care of patients with suspected ACS and without clear ST-elevation infarction 2. The accuracy of AI-powered ECG interpretation in detecting OMI compared to the classical STEMI criteria 3. How infarct size correlates with different ECG readings by AI and (hypothesis generating) if changing the clinical practice could lead to a benefit in patients with suspected OMI.

The 12-lead electrocardiogram (ECG) remains the most immediately accessible and widely used initial diagnostic tool for guiding management in patients with suspected acute coronary syndrome (ACS). Based mainly on the 12-lead ECG, ACS is classically subdivided into ST-elevation myocardial infarction (STEMI) and Non-ST-elevation myocardial infarction (NSTEMI). For patients with a working diagnosis of STEMI, an immediate reperfusion therapy, preferably with a Percutaneous Coronary Intervention (PCI), is recommended. For patients with NSTEMI the timing for an invasive coronary reperfusion is based on risk stratification.To date, the STEMI vs. NSTEMI paradigm prevents some NSTEMI patients with acute occlusion of a coronary artery from receiving emergent PCI, in spite of their known increased mortality compared with NSTEMI without coronary occlusion. Around 24-35% of patients with NSTEMI have total coronary occlusion, referred to as occlusion myocardial infarction (OMI), and could benefit from emergent PCI. These patients often end up receiving reperfusion at a later stage (24-72 hours) into admission, which is thought to be late to salvage ischaemic or infarcted tissue. Many publications since the 2000s describe ECG patterns without ST-segment elevation that signify acute coronary occlusion. However, these ECG signs are subtle and visual inspection of ECG images by clinical experts has been shown to be suboptimal and lead to a high degree of variability in ECG interpretation. A recent study found that machine learning and artificial intelligence (AI) for ECG diagnosis of OMI outperforms practicing clinicians. Also, a novel ECG AI model demonstrated superior accuracy to detect acute OMI when compared to the classical STEMI criteria.This suggests the potential of AI to improve triage of patients with ACS ensuring appropriate and timely referral for immediate PCI. The investigators therefore aim to use these models to analyze ECGs of Patients with NSTE-ACS and to check whether the model outputs OMI or not OMI. Based on that information, the investigators will analyze the time it had taken from admission to intervention (PCI), in order to correlate possible late reperfusions with infarct size of the ventricle. The hypothesis is that a occluded coronary artery will in fact produce a larger infarct size (scar) in the ventricle after longer occlusion times (=reperfusion time), therefore the patients will be dichotomized in early and late intervention patients and analyzed based on their infarct size and outcome, stratified by the OMI diagnosis made by the AI ECG algorithm.