Artificial intelligence and machine learning in myocardial infarction: advances in diagnosis, risk stratification, and clinical decision support

Abstract

Background Cardiovascular disease, and myocardial infarction (MI) in particular, is the leading cause of death worldwide. Electrocardiography, echocardiography, and cardiac biomarkers remain the standard diagnostic tools, but their performance is limited by reader variability and by the need for rapid, highly accurate interpretation in emergency settings. Methods A narrative review was conducted of studies published between 2018 and 2026 in PubMed, Scopus, and Embase examining the integration of artificial intelligence (AI) and machine learning (ML) across diagnostic modalities for MI, with priority given to large or externally validated cohorts, direct comparison with existing clinical pathways, and reported performance metrics. Results AI improved the analysis of cardiac biomarkers, ruling out MI with greater accuracy and efficiency than guideline-recommended pathways, while explainable boosting machines preserved diagnostic performance with added model transparency. In imaging, AI-assisted echocardiography reduced examination time; deep-learning models outperformed conventional quantitative measures in mortality prediction; and contrast-free deep-learning methods detected myocardial scar with high accuracy and specificity relative to late gadolinium enhancement. AI-ECG models showed strong performance in identifying acute coronary occlusion, and the SWEDEHEART-AI neural network outperformed open-source large language models for 1-year mortality prediction. Barriers to implementation include algorithmic bias, limited prospective external validation, model non-interpretability, and fragmented regulatory frameworks. Conclusion AI and ML show substantial promise for improving the diagnosis, risk stratification, and prognostication of MI, but real-world validation, standardised regulation, and explicit ethical safeguards are required before routine clinical implementation. AI should be regarded as a decision-support adjunct that complements, rather than replaces, clinical expertise. Keywords: acute myocardial infarction; artificial intelligence; cardiac biomarkers; cardiovascular disease; deep learning; echocardiography; electrocardiography; explainable machine learning; machine learning; mortality prediction