Fig. 3

The performance on radiomics score derived from prediction model across different cohorts. A, B Comparison of response probability distributions calculated by RF5 model between non-relapse and relapse groups, in the training set and validation set, respectively. Two-sided P values were calculated using the wilcoxon rank-sum test. C, D Confusion matrices revealed predicted outcomes generated by RF5 model, as indicated, in training set and validation set, respectively. Statistical analysis was conducted based on the predictive value and actual value of relapse and non-relapse using the cut off value of radiomics score defined in the training cohort, Chi-square test