Fig. 3: Performance of six algorithms in the prediction of medulloblastoma prognosis in CMR scenario.

a Univariate COX regression feature selection for model construction. b Comparison of Area Under the Receiver Operating Characteristic curves (AUROCs) for predicting medulloblastoma prognosis using six algorithms within different time spans in testing set (left) and external validation set (right). c Receiver operating characteristic (ROC) curves with different algorithms for predicting 5- (left) and 10-year (right) survival predictions in medulloblastoma patients in testing set (n = 59). d ROC curves with Extreme Gradient Boosting algorithm (XGBoost) for predicting 5- (left) and 10-year (right) survival predictions in medulloblastoma patients in external validation set (n = 73). e Calibration plots of overall survival predictions for XGBoost algorithm in testing (left) (n = 59) and external validation set (right) (n = 73). f Decision curves for evaluating the clinical utility (net benefit) of XGBoost algorithm for 5- (left) and 10-year (right) survival predictions (n = 73).