ROC plot of binary logistic regression analysis with backward stepwise selection, using leave-one-out cross-validation method.

Receiver operating characteristics (ROC) curve analysis of binary logistic regression models, using leave-one-out cross-validation method, in differentiating encapsulated from invasive thymomas. The graph shows that the combination of 3D shape analysis and CT features (blue line, AUC, 0.955; 95% CI, 0.935–0.975) has significantly higher discriminating performance in differentiating encapsulated from invasive thymomas compared to clinical and CT features (red line, AUC, 0.666; 95% CI, 0.626–0.707) (difference between AUC values, 0.289; p<0.001). For reference, ROC analysis with 3D shape analysis alone is also demonstrated (green line, AUC, 0.896; 95% CI,0.868–0.923).