search

282857341 / nnFormer

MIT License
798 stars 84 forks source link

Questions of the results on Brain Tumor #65

Open OliverRensu opened 2 years ago

OliverRensu commented 2 years ago

Hi, I run your code but the dice on Brain Tumor dataset is 0.7901 ([0.8388071561002625, 0.6973127182908332, 0.8728819098572764]). Why the results is significantly lower than yours in the paper? Besides, the dice in Table 2 and Table 5 of Brain tumor is different (the Dice score in Table 2 is much higher), is there any difference in settings?

282857341 commented 2 years ago

Brain tumor datasets are evaulated on three classes: ET, TC, WT. But we found the results reported in the paper of UNETR is much lower than the normal value. So we infer that UNETR evaulates on the ED, NET, ET actually. So we report two results which evaulate on ET, TC, WT and ED, NET, ET respectively