Fig. 2

Flowchart of feature selection and radiomics nomogram building. (A) Lesion segmentation and 2D ROI and 3D ROI segmentation; (B) A total of 788 selected features for 2D and 3D ROI respectively; (C) ComBat correction was applied to minimize the potential bias on the results due to different scan protocols of the 5 different CT scanners; (D) Dimension reduction for features selection; (E) Select the optimal algorithm for radiomics model building. The best one was selected by using 5-fold cross-validation in the validation cohort. All the patients were randomly split into 80% for training and the remaining 20% for validation, with 100 iterations; (F) Radiomic nomogram was built on the optimal algorithm. Calibration curves and decision curves were used to evaluate the effectiveness of the radiomics nomogram