Radiomics from CT and MRI for Cardiovascular Risk Assessment: Review and Meta-Analysis

Overview

This comprehensive review and meta-analysis evaluated radiomics applications in cardiac CT and MRI for cardiovascular risk prediction, assessing methodological quality using the novel METRICS score. The pooled diagnostic accuracy of radiomics models demonstrated promising performance, while methodological quality across studies remained generally low to moderate.

Background

Radiomics involves extracting quantitative features from biomedical images to uncover patterns not visible to the naked eye, with growing applications in cardiovascular imaging. Despite its potential to improve diagnosis and risk prediction in cardiac diseases, radiomics research suffers from methodological heterogeneity and low quality. The METRICS score was recently developed to systematically assess radiomics study quality, yet no prior review has applied it to cardiac imaging. This study aimed to fill that gap by reviewing radiomics studies in cardiac CT and MRI and performing a meta-analysis focused on cardiovascular event prediction.

Data Highlights

Key Findings

• Radiomics models from cardiac CT and MRI show promising diagnostic accuracy for predicting cardiovascular events, as demonstrated by pooled AUC values.
• Methodological quality of radiomics studies in cardiac imaging remains suboptimal, with many studies scoring low to moderate on the METRICS scale.
• The METRICS score provides a structured and systematic approach to evaluate radiomics study quality, addressing previous limitations of the Radiomics Quality Score (RQS).
• Significant heterogeneity exists in study aims, methodologies, and reporting standards across radiomics research in cardiovascular imaging.
• Publication bias was assessed and adjusted using funnel plots and trim-and-fill methods, supporting the robustness of meta-analytic results.
• Sensitivity analyses confirmed the stability of pooled diagnostic performance estimates despite individual study variability.

Clinical Implications

Radiomics extracted from cardiac CT and MRI holds potential as a non-invasive tool for cardiovascular risk stratification, potentially enhancing clinical decision-making. However, the current low to moderate methodological quality of studies highlights the need for standardized protocols and rigorous study designs to facilitate clinical translation. Adoption of quality assessment tools like METRICS can guide researchers in improving study robustness and reproducibility.

Conclusion

Radiomics applications in cardiac imaging demonstrate encouraging diagnostic accuracy for cardiovascular risk prediction, yet methodological limitations persist. Systematic quality assessment using METRICS underscores the need for improved study design to realize radiomics' full clinical potential.

References

1. Ponsiglione et al 2021 -- Systematic review of radiomics studies in cardiac CT and MRI
2. EuSoMII Radiomics Auditing Group 2024 -- Introduction of METRICS score for radiomics quality assessment