(TMI 2024)Physics-informed Score-based Diffusion Model for Limited-angle Reconstruction of Cardiac Computed Tomography [PyTorch]

Official PyTorch implementation of PSDM. Code modified from Diffusion-MBIR. This is just for the simulation purpose. The code for real experiment will update soon~

Getting started

• Install dependencies

python = 3.7
torch = 1.13.1
scipy = 1.7.3
astra-toolbox
Operator Discretization Library (ODL)
sporco,tqdm, ninja,ml_collections

PSDM reconstruction

Once you have the pre-trained weights and the test data set up properly, you may run the following scripts. Modify the parameters in the python scripts directly to change experimental settings.

conda activate your_conda
python imnverse_me.py

Training

You may train the diffusion model with your own data by using e.g.

bash train_AAPM256.sh

You can modify the training config with the --config flag.

Citation

If you find our work interesting, please consider citing

@article{han2024physics,
  title={Physics-informed Score-based Diffusion Model for Limited-angle Reconstruction of Cardiac Computed Tomography},
  author={Han, Shuo and Xu, Yongshun and Wang, Dayang and Morovati, Bahareh and Zhou, Li and Maltz, Jonathan S and Wang, Ge and Yu, Hengyong},
  journal={arXiv preprint arXiv:2405.14770},
  year={2024}
}