Figure 2

ROC curves for the ML classifiers, pathologists, and hybrid models on the WCM test data. (A) compares the model performance of the single-scale ensembles and the multi-scale ensemble. (MSE). The performance of the semiquantitative predictions of two expert neuropathologists and the two-pathologist averaged consensus are compared in B). (C) compares the predictions of the top-performing neuropathologist with the MSE, and the hybrid model generated by naïve averaging of pathologist and MSE predictions.