Table 5.
Summary of the performance of models trained and evaluated on data with bias field on MS dataset.
MS dataset-test set (with bias field) | ||||
---|---|---|---|---|
CNNMS-BF | CNNMS-BF-DA | |||
Tissue | DC(P50) | DC(P10) | DC(P50) | DC(P10) |
WM | 0.950 | 0.920 | 0.945 | 0.914 |
GM | 0.914 | 0.877 | 0.903 | 0.859 |
LV | 0.941 | 0.854 | 0.845 | 0.554 |
Tha | 0.937 | 0.911 | 0.937 | 0.896 |
HC | 0.865 | 0.842 | 0.858 | 0.827 |
CdN | 0.915 | 0.880 | 0.910 | 0.865 |
Pu | 0.903 | 0.885 | 0.911 | 0.891 |
GP | 0.855 | 0.803 | 0.854 | 0.807 |
ALL | 0.910 | 0.872 | 0.895 | 0.827 |
DC, Dice scores. Highlighted results indicate that median values are larger (P50: Wilcoxon, p < 0.05) or variances are lower (P10: Levene, p < 0.05).