Abstract 4369224: Artificial Intelligence ECG-Extracted Features Predict Microvascular Obstruction in ST-segment Elevation Myocardial Infarction

Background: Microvascular obstruction (MVO) is a critical predictor of adverse outcomes including mortality and the development of heart failure following ST-segment Elevation Myocardial Infarction (STEMI). Although cardiac magnetic resonance imaging (CMR) remains the gold standard for detection and quantification of MVO, it is resource intensive and has limited availabity. Artificial intelligence (AI)-extracted electrocardiographic (ECG) features may offer a rapid, non-invasive tool for predicting MVO risk in the hospital setting both pre- and post-percutaneous coronary intervention (PCI). Additionally, it may prove valuable in identifying high-risk patients in STEMI clinical trials when CMR is not available. Objective: To evaluate the feasibility and performance of an AI-enhanced ECG approach to predict MVO in anterior STEMI patients using pre- and post-PCI ECGs with MVO measured by CMR. Methods: Paients with anterior STEMI who underwent primary PCI at a single Institution were retrospectively analyzed. Each patient had both an admission ECG obtained in the ER and a post-PCI ECG. All patients underwent CMR 1-2 days following successful PCI for measurement of infarct size and MVO. A logistic regression model was developed using AI-extracted ECG features from both timepoints. Model performance was assessed by specificity, sensitivity and positive predictive value (PPV). Feature importance was evaluated using SHAP values. Results: Among 99 patients, 50 (50%) had MVO confirmed by CMR. The AI ECG-based logistic regression model achieved an AUC of 0.83, with a specificity of 94%, sensitivity of 60% and a PPV of 86% (Fig 1A). The top five most predictive ECG features of MVO identified by SHAP included (Fig 1B): 1.)High Active Occlusion score on Post-PCI ECG; 2.)Low/absent reperfusion score on post-PCI ECG; 3.)Extent of ST-elevation on pre-PCI ECG; 4.)Small change in active occlusion score between pre- and post-PCI ECGs; 5.)Reduced EF on post-PCI ECG. Conclusion: An AI-enhanced ECG approach utilizing pre-and post-PCI ECGs demonstrate promising predictive performance for identifying MVO in anterior STEMI patients. This approach could enable real-time risk stratification in acute settings and inform theapeutic decisions.