- Prepare the data using the 'dataset' instructions below.
- Convert nii.gz to .png files using the preprocessing instructions specified in Section 4.3 of the paper.
- We experimented 3 image synthesis models as shown below:
- Follow the data preparation procedures specified by each image translation model and train the image translation models. We use the default experiment setup to train the models.
- Repeat step 1. to prepare the testing data, and run the inference using the trained image synthesis model to generate synthetic CTs.
- Gather the 2D synthetic CT images generated by image-to-image translation models and reconstruct them into a 3D volume.
- Use TotalSegmentator to run the inference on the synthetic 3D volume to obtain the segmentation results.
- TotalSegmentator (https://github.com/wasserth/TotalSegmentator)
-
Since we don't own the datasets used in the study, you need to download them. For training, we used Gold Atlas - Male Pelvis dataset while the CPTAC-UCEC for the testing. Please see the detailed information below to obtain the datasets.
-
Training: Gold Atlas - Male Pelvis dataset, which can be downloaded here: https://zenodo.org/records/583096
-
Testing:
CPTAC-UCEC Dataset, which can be downloaded here: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=19039602 . We used the T2 sequences as shown below. Then since the raw image files were in DICOM format, we used itk-snap to convert them to NIFTI format.├── CPTAC-UCEC\C3N-00858\01-08-2000-NA-MIEDNICA-65907\4.000000-t2bltsetra-76193 ├── CPTAC-UCEC\C3N-00860\01-22-2000-NA-MIEDNICA-33628\4.000000-t2bltsetra-29838 ├── CPTAC-UCEC\C3N-01001\02-23-2000-NA-MIEDNICA-97539\4.000000-t2bltsetra-55931 ├── CPTAC-UCEC\C3N-01003\02-19-2000-NA-MIEDNICA-66632\4.000000-t2bltsetra-09021 ├── CPTAC-UCEC\C3N-01007\04-01-2000-NA-MIEDNICA-78405\4.000000-t2bltsetra-13404 ├── CPTAC-UCEC\C3N-01172\05-04-2000-NA-MIEDNICA-78744\4.000000-t2bltsetra-03604 ├── CPTAC-UCEC\C3N-01341\05-09-2000-NA-BODYMIEDNICA-18754\5.000000-t2bltsetra-20976 ├── CPTAC-UCEC\C3N-01342\05-10-2000-NA-MIEDNICA-52074\4.000000-t2bltsetra-75036 ├── CPTAC-UCEC\C3N-01346\05-26-2000-NA-BODYMIEDNICA-60801\3.000000-t2bltsetra-10742 ├── CPTAC-UCEC\C3N-01761\06-08-2000-NA-MIEDNICA-08891\4.000000-t2bltsetra-02893 ├── CPTAC-UCEC\C3N-01763\06-01-2000-NA-MIEDNICA-68213\4.000000-t2bltsetra-02370 ├── CPTAC-UCEC\C3N-01764\06-06-2000-NA-BODYMIEDNICA-75164\4.000000-t2bltsetra-30211 ├── CPTAC-UCEC\C3N-01765\06-24-2000-NA-MIEDNICA-13084\4.000000-t2bltsetra-64390 ├── CPTAC-UCEC\C3N-01871\05-31-2000-NA-MIEDNICA-91535\4.000000-t2bltsetra-32087 ├── CPTAC-UCEC\C3N-01873\07-01-2000-NA-MIEDNICA-08363\4.000000-t2bltsetra-62871 ├── CPTAC-UCEC\C3N-01875\07-15-2000-NA-MIEDNICA-75940\4.000000-t2bltsetra-17941 ├── CPTAC-UCEC\C3N-01876\07-27-2000-NA-MIEDNICA-61790\4.000000-t2bltsetra-55722 ├── CPTAC-UCEC\C3N-01877\08-10-2000-NA-MIEDNICA-26947\4.000000-t2bltsetra-14823 ├── CPTAC-UCEC\C3N-01878\08-03-2000-NA-MIEDNICA-17690\4.000000-t2bltsetra-13491 ├── CPTAC-UCEC\C3N-01879\08-16-2000-NA-MIEDNICA-16519\4.000000-t2bltsetra-62254 ├── CPTAC-UCEC\C3N-01880\09-02-2000-NA-MIEDNICA-71707\4.000000-t2bltsetra-26489 ├── CPTAC-UCEC\C3N-02595\12-08-2000-NA-BODYMIEDNICA-36410\3.000000-t2bltsetra-18127 ├── CPTAC-UCEC\C3N-02631\12-09-2000-NA-MIEDNICA-37114\4.000000-t2bltsetra-07292 ├── CPTAC-UCEC\C3N-02632\11-08-2000-NA-MIEDNICA-08519\4.000000-t2bltsetra-40678 ├── CPTAC-UCEC\C3N-02639\01-04-2001-NA-MIEDNICA-14771\4.000000-t2bltsetra-00520 ├── CPTAC-UCEC\C3N-02678\12-29-2000-NA-BODYMIEDNICA-71292\3.000000-t2bltsetra-96765 ├── CPTAC-UCEC\C3N-02976\01-03-2001-NA-MIEDNICA-34482\4.000000-t2bltsetra-22047 ├── CPTAC-UCEC\C3N-02978\12-28-2000-NA-MIEDNICA-99235\4.000000-t2bltsetra-71537 ├── CPTAC-UCEC\C3N-02979\01-02-2001-NA-BODYMIEDNICA-43265\3.000000-t2bltsetra-51072 ├── CPTAC-UCEC\C3N-03417\04-05-2001-NA-MIEDNICA-82370\4.000000-t2bltsetra-39283
If you find our work is useful for your research, please consider citing
@article{zhuang2024segmentation,
title={Segmentation of pelvic structures in T2 MRI via MR-to-CT synthesis},
author={Zhuang, Yan and Mathai, Tejas Sudharshan and Mukherjee, Pritam and Summers, Ronald M},
journal={Computerized Medical Imaging and Graphics},
pages={102335},
year={2024},
publisher={Elsevier}
}