Coronary Artery Disease Assessment with Multimodal Machine Learning: A Systematic Review

Abstract

Coronary artery disease (CAD) is the most common cause of mortality worldwide, and its management requires the integration of information from diverse data sources. While Machine Learning models have shown their potential in supporting patient care across different healthcare fields, the multimodal nature of clinical data remains a considerable challenge to the use of these models. A systematic review was conducted (PROSPERO: CRD420251081494) and 119 eligible studies were identified. The review found that ML methods have been employed across multiple CAD-related tasks, including risk factor identification, disease detection, severity stratification/prognosis prediction, and treatment guidance. While one third of the studies employed multimodal data, the methodologies for multimodal data integration were frequently simplistic, often relying on manual feature extraction. This review is the first to systematically assess multimodal ML strategies in CAD. The findings point to a need for more robust data integration and advanced multimodal fusion techniques and standardized evaluation to translate ML innovations into better CAD patient care.