Influence of Different Segmentations on the Diagnostic Performance of Pericoronary Adipose Tissue

Figure 1. Representative CCTA image of PCAT around LAD. A hemodynamically significant stenosis (FFR = 0.73) in the middle segment of LAD. PCAT CT attenuation was defined as the mean CT attenuated value within a radial distance equal to the diameter of the vessel. The length of the lesion-based PCAT was defined from the proximal to the distal shoulder of the lesion (A), the vessel-based PCAT was focused on the proximal 40-mm segments of all three major epicardial coronary vessels (10 to 50 mm from right coronary artery ostium) (B). LAD, left anterior descending artery; PCAT, pericoronary adipose tissue.

Figure 2. Correlation and Bland-Altman plots of vessel-based and lesion-based PCAT CT attenuation as clarified by invasive FFR. Pearson correlation analyses show the correlation vessel-based and lesion-based PCAT CT attenuation in all vessels (A), vessels with FFR > 0.80 (B) and vessels with FFR ≤ 0.80 (C), respectively. Bland-Altman plots show PCAT CT attenuation between vessel-based and lesion-based in all vessels (D), vessels with FFR >0.80 (E) and vessels with FFR ≤ 0.80 (F). The red dotted lines represent the mean difference (bias), and the blue dotted lines represent the 95% limits of agreement. FFR, fractional flow reserve; PCAT, pericoronary adipose tissue.

Figure 3. Correlation and Bland-Altman plots of vessel-based and lesion-based PCAT CT attenuation in different coronary arteries. Pearson correlation analyses show the correlation vessel-based and lesion-based PCAT CT attenuation in RCA (A), LAD (B) and LCX (C), respectively. Bland-Altman plots show PCAT CT attenuation between vessel-based and lesion-based in RCA (D), LAD (E) and LCX (F). The red dotted lines represent the mean difference (bias), and the blue dotted lines represent the 95% limits of agreement. LAD, left anterior descending artery; LCX, left circumflex artery; PCAT, pericoronary adipose tissue; RCA, right coronary artery.

Figure 4. ROC curve analysis of PCAT CT attenuation measured from lesion- and vessel-based segmentations for identifying functionally significant coronary stenosis. Two segmentations had similar AUC (0.547 vs. 0.524, p = 0.814). AUC, area under the curve; ROC, receiver operating characteristic curve.

Figure 5. Manhattan plots demonstrated the strength of association [–log10 (P-value)] of univariate logistic regression] comparing PCAT radiomics features derived from lesion-based (A) and vessel-based (B) PCAT segmentation against the hemodynamic significance of coronary stenosis. Solid points above the horizontal cutoff line denote features qualified for univariate logistic regression and LASSO-logistic regression, and their corresponding feature names are displayed. LASSO, least absolute shrinkage and selection operator; PCAT, pericoronary adipose tissue.

Figure 6. ROC curves indicated comparable diagnostic performance between vessel-based and lesion-based PCAT radiomics models for identifying functionally significant coronary stenosis. PCAT, pericoronary adipose tissue; ROC, receiver operating characteristic.