Fig. 5

A five-marker three-class stratification model for bladder cancer risk stratification. a Performance features of top individual markers for risk stratification in Cohort 1 from single marker analysis; average of balanced accuracies of the three groups (where balanced accuracy was calculated as ½ of sum of sensitivity and specificity of each group), overall AUCs (where AUC referred to area under ROC generated by sensitivity and specificity summing up of true positive, false positive, true negative and false negative of each class based on the micro-average method) and overall accuracies were expressed as mean with 95% CI in 100 splits of train-test sampling. b ROC curves of stratification model in Cohorts 1 and 2; AUC referred to area under ROC generated by sensitivity and specificity summing up of true positive, false positive, true negative and false negative of each class based on the micro-average method; c distributions of the three-class probabilities from the stratification model for samples in Cohorts 1 and 2; each sample was depicted by the three coordinates representing the probabilities of the respective non-BC, LMR-NMIBC or HR-NMIBC + MIBC group