About ACDC dataset

[Heonjin33]

Thanks for your work!

I have a question about the dice for ACDC dataset. In your experiment, I found that your initial dice is: 62.32(LV), 49.04(Myo), 66.86(RV), 68.87(Overall) which is absolutely different from FSDiffReg: 58.5(LV), 35.7(Myo), 74.1(RV), 65.5(Overall) and DiffuseMorph: 64.2(Overall)

The first time I used the initial dataset which was downloaded from ACDC dataset, and executed the codes from FSDiffReg, I couldn't get same dice to FSDiffReg. Then, from the issue: issue in FSDiffReg, I did the same preprocess. I still can't get the same dice to FSDiffReg, but I got the same dice to your work.

I wonder if the dataset had been modified or updated, which causes this situation. Could anyone understand my problem and give me some help or ideas? I would greatly appreciate!

[xuexue49]

maybe they use a 3D-Dice