search

cwmok / Conditional_LapIRN

Conditional Deformable Image Registration with Convolutional Neural Network
MIT License
80 stars 16 forks source link

MIND-SSC implementation for lung/abdominal MR-CT #13

Closed annareithmeir closed 1 year ago

annareithmeir commented 1 year ago

Which MIND-SSC features implementation are you using for the lung/abdominal MR-CT dataset?

And in your slides of the miccai 2021 oral are you training this dataset unsupervised or semi-supervised? If the latter, do you remember if the unsupervised version gives good results too?

Thanks!

cwmok commented 1 year ago

Hi @annareithmeir,

I use the MIND-SSC implementation at https://github.com/multimodallearning/convexAdam/blob/main/convex_adam_utils.py

For the Learn2Reg 2021 challenge, I use semi-supervised loss (with dice loss). I didn't try the unsupervised version for the MR-CT task. However, I did try the unsupervised version for abdominal CT-CT registration. It shows ~10% drop in Dice. The result is in https://arxiv.org/abs/2306.16997.