This is a copy of package for medical image synthesis work with LRes-ResUnet and GAN (wgan-gp) in pytorch framework, which is a simple extension of our paper Medical Image Synthesis with Deep Convolutional Adversarial Networks. You are also welcome to visit our Tensorflow version through this link: https://github.com/ginobilinie/medSynthesis
The main entrance for the code is runCTRecon.py or runCTRecon3d.py (currently, the 2d/2.5d version is fine to run, and the discriminator for 3d version currently only support BCE loss since I suggest you use W-distance (WGAN-GP) since it is easier to tune the hyper-parameters for this one).
I suppose you have installed:
python 2.x (e.g., 2.7.x; for python 3.x, change some codes: .next() to ._next_(); xrange()->range())
pytorch (>=0.3.0)
simpleITK
numpy
Steps to run the code:
If it is helpful to your work, please cite the papers:
@inproceedings{nie2017medical, title={Medical image synthesis with context-aware generative adversarial networks}, author={Nie, Dong and Trullo, Roger and Lian, Jun and Petitjean, Caroline and Ruan, Su and Wang, Qian and Shen, Dinggang}, booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention}, pages={417--425}, year={2017}, organization={Springer} }
@article{nie2018medical, title={Medical Image Synthesis with Deep Convolutional Adversarial Networks}, author={Nie, Dong and Trullo, Roger and Lian, Jun and Wang, Li and Petitjean, Caroline and Ruan, Su and Wang, Qian and Shen, Dinggang}, journal={IEEE Transactions on Biomedical Engineering}, year={2018}, publisher={IEEE} }
BTW, you can download a real medical image synthesis dataset for reconstructing standard-dose PET from low-dose PET via this link: https://www.aapm.org/GrandChallenge/LowDoseCT/
Also, there are some MRI synthesis datasets available: http://brain-development.org/ixi-dataset/
Tumor prediction: https://www.med.upenn.edu/sbia/brats2018/data.html
fastMRI: https://fastmri.med.nyu.edu/
ISLES2015: http://www.isles-challenge.org/ISLES2015/
If you're interested in it, you can send me a copy of your data (for example, brain MRI), and I'll inference the CT and send a copy of predicted CT to you. My email is dongnie.at.cs.unc.edu.
medSynthesis is released under the MIT License (refer to the LICENSE file for details).