CiSeg: Unsupervised Cross-Modality Adaptation for 3D Medical Image Segmentation via Causal Intervention
-
MMWHS dataset：多模态全心分割挑战
-
Abdominal CT: https://www.synapse.org/#!Synapse:syn3193805/wiki/217789
-
CHAOS MRI:Description - CHAOS - Grand Challenge
Our preprocessed data and weights are available at the following link.
[Baidu Netdisk](通过网盘分享的文件：cardiac_abdominal 链接: https://pan.baidu.com/s/1yIainEESJZnIiPLHm4u16Q 提取码: 3bxh --来自百度网盘超级会员v6的分享)
deeplabv2 pre-training weights
The file structure should be:
data
cardiac
tr_ct
image
ct_train_1001_image.nii.gz
....
label
ct_train_1001_label.nii.gz
...
tr_ct_slice
ct_train_1001_image_slice_0.h5
ct_train_1001_image_slice_1.h5
...
tr_mr
image
mr_train_1001_image.nii.gz
...
label
mr_train_1001_label.nii.gz
...
tr_ct_slice
mr_train_1001_image_slice_0.h5
mr_train_1001_image_slice_1.h5
...
val_ct
image
...
label
...
val_ct_slice
ct_train_1003_image_norm.h5
...
val_mr
image
...
label
...
val_mr_slice
mr_train_1007_image_norm.h5
...
tr_ct_slices.list
tr_mr_slices.list
val_ct_slices.list
val_mr_slices.list
abdominal
tr_ct
image
label
tr_ct_slice
tr_mr
image
label
tr_ct_slice
val_ct
image
label
val_ct_slice
val_mr
image
label
val_mr_slice
tr_ct_slices.list
tr_mr_slices.list
val_ct_slices.list
val_mr_slices.list
python dataloaders/pre_mmwhs.py
python dataloaders/nii2h5.py
Or please refer to GenericSSL.
python dataloaders/pre_abdominal.py
python dataloaders/nii2h5.py
eg：Cardiac CT
python train_fully_supervised.py
python test.py
eg：Cardiac MRI2CT
python train_CiSeg.py
python test_CiSeg.py
AdaOutput、AdvEnt、 CycleGAN、 SAFAv2、MPSCL、
SE-ASA、 CPCL、MIC、 ASC、MAAL、DCLPS
This code is mainly based on SSL4MIS and Learning-Debiased-Disentangled.