Skip to main content
. 2023 Jun 27;15(13):3369. doi: 10.3390/cancers15133369

Table 5.

Performance evaluation of the model trained on the largest slice using entire brain images and data augmentation.

Mean 3-Fold CV Ensemble 3-Fold CV
Train Validation Test Test (95% CI)
LGG vs. HGG
 Loss 0.451 0.474 0.459
 Accuracy 0.793 0.797 0.797 0.805 (0.759–0.851)
 AUC-ROC 0.874 0.864 0.866 0.878 (0.840–0.916)
 Precision
  LGG 0.600 0.611 0.606 0.624 (0.567–0.680)
  HGG 0.917 0.911 0.913 0.906 (0.872–0.940)
 Recall
  LGG 0.822 0.804 0.809 0.788 (0.740–0.835)
  HGG 0.781 0.794 0.792 0.812 (0.766–0.857)
 F1
  LGG 0.693 0.694 0.701 0.696 (0.642–0.750)
  HGG 0.843 0.848 0.856 0.856 (0.815–0.897)
Grade (2/3/4)
  Loss 0.867 0.837 0.802
  Accuracy 0.729 0.751 0.742 0.752 (0.701–0.803)
    G.2 0.801 0.793 0.794 0.802 (0.755–0.849)
    G.3 0.865 0.889 0.892 0.892 (0.856–0.929)
    G.4 0.793 0.819 0.798 0.809 (0.763–0.856)
  AUC-ROC 0.776 0.788 0.803 0.816 (0.770–0.861)
    G.2 0.838 0.849 0.862 0.872 (0.833–0.911)
    G.3 0.865 0.670 0.697 0.713 (0.660–0.766)
    G.4 0.793 0.844 0.850 0.861 (0.821–0.902)
  Precision
    G.2 0.427 0.422 0.434 0.448 (0.390–0.507)
    G.3 0.246 0.000 0.000 0.000 (0.000–0.000)
    G.4 0.869 0.880 0.881 0.890 (0.853–0.927)
  Recall
    G.2 0.655 0.726 0.812 0.848 (0.806–0.890)
    G.3 0.099 0.000 0.000 0.000 (0.000–0.000)
    G.4 0.842 0.871 0.836 0.842 (0.799–0.885)
  Specificity
    G.2 0.829 0.806 0.790 0.793 (0.745–0.841)
    G.3 0.960 1.000 1.000 1.000 (1.000–1.000)
    G.4 0.662 0.683 0.697 0.724 (0.671–0.776)
  F1
    G.2 0.517 0.532 0.520 0.586 (0.529–0.644)
    G.3 0.139 0.000 0.000 0.000 (0.000–0.000)
    G.4 0.662 0.875 0.872 0.865 (0.825–0.905)