Abstract
OBJECTIVES: To conduct a comprehensive systematic review of the studies applying radiomics to CT and MRI for the evaluation of cardiac disease, and to perform a meta-analysis of their diagnostic accuracy, focused on cardiovascular events prediction. A secondary aim was to assess the methodological quality of cardiac imaging radiomics studies using the METRICS score. MATERIALS AND METHODS: Four investigators searched multiple medical literature archives (Scopus, Web of Science, and PubMed). The search was conducted from February 7th, 2021, to March 10th, 2025. Papers were also screened to identify studies for the prediction of cardiovascular events, defined as the occurrence of major cardiovascular events or myocardial ischemia. Methodological quality was assessed by using the METRICS tool. Diagnostic accuracy was estimated with pooled area under the curve (AUC). RESULTS: A total of 202 studies were included in the final analysis. Seventeen papers were identified for the meta-analysis, of which 9 were considered eligible for analysis. 111 papers (55%) had CT as the imaging modality, and 91 (45%) papers had MRI. Overall, the average METRICS total score was 54.52% ± 15.89%. Meta-analysis showed pooled AUC of 0.81 (95% CI: 0.75-0.87), with a high level of heterogeneity (I² = 83.4%, τ² = 0.0068). Egger's test for funnel plot asymmetry was statistically significant (z = -2.39, p = 0.017), suggesting potential publication bias. CONCLUSION: Radiomics in cardiac imaging holds potential, showing moderate quality and relatively high cumulative performance for the prediction of cardiovascular events. KEY POINTS: Question What is the current methodological quality and pooled diagnostic performance of cardiovascular radiomics for predicting clinical events, based on a meta-analysis? Findings The average METRICS quality score was 54.52%. A meta-analysis showed a pooled AUC of 0.81 for event prediction, but with high heterogeneity and publication bias. Clinical relevance Assessing radiomics research methodological quality is crucial to enhance reproducibility and clinical applicability of radiomics pipelines. The evaluation of cumulative evidence for cardiovascular events prediction may guide clinical translation and future study design.