Fig. 4From: Novel prognostication of patients with spinal and pelvic chondrosarcoma using deep survival neural networksPerformance evaluation of RED_SNN using test data set. a ROC curves to evaluate the prediction with test data set. The test data were analyzed by pre-trained RED_SNN and its output (expected survival probability) was compared to the real survival of the test set at each time point. b Calibration curves of RED_SNN model to predict the survival rate of the test set. The test data were analyzed using pre-trained RED_SNN, and the test patients were equally divided into seven subgroups according to the model’s predicted survival probability at five yearsBack to article page