Fig. 3

a–d The learning curves for four different radiomics models. The red and green curves represent the trend of the score with the increase in sample size in training and cross-validation data, respectively. The training and cross-validation scores of the logistic regression (LR) and SVM models were higher than those of the Naive Bayes (NB) model. The gap between the training and cross-validation scores of the LR or SVM models was smaller than that of the Random Forest (RF) model. e–h The receiver operating characteristic curves for four different radiomics models. Each light-colored curve represents each of the tenfold cross-validations (fold 0 to 9), and the dark blue curve represents their mean; the red and green curves represent internal and external validation sets, respectively. The LR model and SVM model had similar AUCs in the training set, but the LR model outperformed the SVM model in both the internal validation set and the external validation set. The RF model had the highest AUC in the training set but had low AUC in the external validation. The NB model had the lowest AUCs in all sets