Integrative Deep Learning Models for Predicting Myocardial Infarction from EHR and Imaging Data

Background: Early and accurate prediction of myocardial infarction (MI) remains a cornerstone of preventive cardiology. Traditional risk scores, while useful, rely on limited clinical variables and often fail to account for the complexity and heterogeneity of patient data. Integrative deep learning approaches that combine electronic health records (EHR) with imaging data offer a novel solution by capturing high-dimensional, multimodal features reflective of both structural abnormalities and systemic risk profiles. These models hold promise for enhancing risk stratification, facilitating timely intervention, and improving patient outcomes. Methods and Results: This study developed and validated a series of deep learning architectures integrating longitudinal EHR data—including demographics, laboratory values, medications, and comorbidities—with echocardiographic and coronary CT imaging features. Using a dataset of over 100,000 patients from multiple health systems, convolutional neural networks (CNNs) and transformer-based models were trained to predict incident MI within a 1-year window. Model performance was evaluated using ROC-AUC, precision-recall curves, and decision curve analysis, with external validation on an independent hospital cohort. The best-performing model achieved an AUC of 0.93, significantly outperforming traditional models such as the Framingham Risk Score and ASCVD calculator. Conclusion: Integrative deep learning models that combine EHR and imaging data provide a powerful and scalable framework for early prediction of myocardial infarction. By leveraging multimodal data sources and advanced neural architectures, these tools can augment clinical decision-making and enable personalized, preemptive cardiovascular care. Future efforts should focus on real-time deployment, clinician interpretability, and regulatory pathways for clinical integration. Conclusion: Integrative deep learning models that combine EHR and imaging data provide a powerful and scalable framework for early prediction of myocardial infarction.