Radiomic phenotype of epicardial adipose tissue derived from coronary artery calcium score predicts myocardial ischemia.

Abstract

Purpose: To investigate the radiomic phenotype of epicardial adipose tissue (EAT) derived from coronary artery calcium score (CACS) and its predictive value for myocardial ischemia.

Materials and methods: This retrospective study included patients with angina and intermediate-to-high pre-test probability of coronary artery disease who underwent CACS, dynamic CT myocardial perfusion imaging (CT-MPI) and coronary CT angiography (CCTA). All image acquisitions were performed with third generation dual source CT. Radiomic features of EAT derived from CACS were extracted. EAT volume, EAT density, Coronary Artery Disease-Reporting and Data System (CAD-RADS) grades, CACS, and clinical characteristics were recorded. The diagnostic abilities of CT-derived parameters, clinical + CACS model, the EAT radiomic model, and combined model for identification of myocardial ischemia (defined as quantitative myocardial blood flow of less than 100 mL/100 mL/min) were evaluated.

Results: A total of 555 patients from two hospitals were included and divided into training set and external validation set separately. The EAT radiomic model was found to have a larger area under the curve (AUC) (0.840 for training set, 0.838 for validation set) than other CT-derived parameters and the clinical + CACS model for predicting myocardial ischemia (all p < 0.05). The overall diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of radiomic model in validation set were 65%, 80%, 59%, 41% and 89%, respectively.

Conclusion: The EAT radiomic model demonstrated superior diagnostic performance over clinical + CACS model and other CT-derived parameters in discriminating myocardial ischemia with highest sensitivity and NPV. Nevertheless, the PPV of the EAT radiomic model was found to be low.