Fig. 3

Importance scores of predictors selected from 216 features, including 10 tumor locations and 206 radiomics features: (A) The most important predictor was temporal lobe involvement (location), followed by high dependence high grey level emphasis (texture), elongation (shape), area density (axis-aligned bounding box, shape feature), information correlation 1 (Grey Level Co-occurrence Based Features), midbrain (location), normalized inverse difference (texture) and intensity range (intensity feature); (B) ROC curves depicting the validation performance of models based on tumor location alone, radiomic features alone, and the combination of both tumor location and radiomic features. Receiver operating characteristic curve (ROC), area under curve (AUC), and accuracy (ACC)