Epicardial Adipose Tissue Volume Assessment in the General Population and CAD-RADS 2.0 Score Correlation Using Dual Source Cardiac CT

Figure 1. Enrollment flowchart of the study.

Figure 2. (A–C) Non-contrast cardiac CT axial views at three different anatomical levels of the same patients (EAT volume: 414 cm3). (A) shows pericardial contours drawn at the most superior slice, at the bifurcation of the main pulmonary artery; (B) pericardial contours at the medial slice and (C) at the most inferior slice, where the posterior descending coronary artery in the inferior atrioventricular groove is last seen. The segmentation process of the epicardial adipose tissue (green areas) was completed using the quantification software.

Figure 3. (A–C) show basal cardiac axial scans obtained through non-contrast computed tomography (NCCT) of individuals with varying epicardial adipose tissue (EAT) volumes. The images represent three distinct categories: low EAT volume (35 cm3) (A), average EAT volume (125 cm3) (B), and high EAT volume (414 cm3) (C).

Figure 4. Scatter plot depicting the distribution of epicardial adipose tissue (EAT) volume frequency (y-axis) in women and men, stratified by age (x-axis). This visualization highlights the relationship between age and EAT volume for each sex, providing a comparative perspective on how EAT accumulation differs between men and women across various age groups.

Figure 5. Graphic in the figure illustrates the coefficients from the multivariate logistic regression model along with their 95% confidence intervals for the variables: log-transformed EAT volume, age, and coronary artery calcium (CAC). The log-transformed EAT volume shows a strong positive association with coronary artery disease (CAD). The confidence interval does not cross zero, indicating that higher levels of epicardial fat significantly increase the likelihood of CAD.

Figure 6. Comparison of ROC curves for logistic regression models with and without CAD-RADS. The model including CAD-RADS (AUC = 1.0) shows perfect discrimination, while the model excluding CAD-RADS (AUC = 0.799) retains good predictive power but with reduced accuracy.