Predicting major adverse cardiac events using radiomics models on coronary computed tomography angiography: A...

Cardiovascular diseases remain a leading cause of morbidity and mortality worldwide, necessitating advanced diagnostic tools for risk stratification. Coronary computed tomography angiography (CCTA) combined with radiomics-a computational method extracting quantitative features from medical images-has emerged as a promising approach to predict major adverse cardiac events (MACEs) in patients who underwent CCTA for suspected coronary lesions.

This systematic review and meta-analysis aimed to evaluate the diagnostic performance of radiomics-based models derived from CCTA for predicting MACEs.

We conducted a comprehensive literature search across PubMed, Embase, and Cochrane Central. The main outcome was pooled MACEs predictability estimates, measured by pooled area under the curve (AUC), hazard ratios (HRs) and C-statistics. Subgroup analyses explored performance by radiomic features (e.g., pericoronary adipose tissue [PCAT]) and patient populations. Methodological quality was assessed by using the METRICS tool.

Eleven studies meeting inclusion criteria were analyzed. The pooled AUC for radiomics models was 0.800 (95% CI: 0.732-0.868; I² = 75.1%). PCAT-based models showed lower consistency (AUC: 0.777, I² = 80.7%) compared to non-PCAT models (AUC: 0.859, I² = 0%). Subgroup analyses revealed AUCs of 0.754 for coronary artery disease (CAD), 0.901 for suspected/confirmed CAD. Univariate HR was 2.54 (95% CI: 2.00-3.24), while multivariate HR was 1.34 (95% CI: 1.04-1.72). Overall, the average METRICS total score was 70.32% ± 14.20%.

Radiomics-based CCTA models demonstrate robust performance for MACE prediction, with variability tied to feature selection and patient populations. These findings highlight radiomics' potential to enhance risk stratification and guide personalized interventions.