This repository contains the implementation of "Learning to Distill Global Representation for Sparse-View CT" in ICCV2023. It includes two main components:
(1) simulator protocol for Sparse-View CT: A simulator protocol for sparse-view CT during the training process in CUDA. Designed as an easy-to-use wrapper, it is compatible with various networks and datasets.
(2) Fourier network and distillation framework: An effective image-domain-only method for sparse-view CT reconstruction.
- clean code and init commit
The DeepLesion dataset is available at DeepLesion, and the AAPM-Myo dataset can be downloaded from CT Clinical Innovation Center.
To set up the environment, please refer to requirements.txt. Notably, we utilize torch radon for efficient projection and forward-back projection. Follow these steps for installation, ensuring compatibility with the specified versions of CUDA and Torch:
git clone https://github.com/matteo-ronchetti/torch-radon.git
cd torch-radon
python setup.py installPlease refer to RUN.sh in the src directory for examples of training and testing
If you find our work and code helpful, please kindly cite our paper:
@InProceedings{GloReDi_2022,
author = {Li, Zilong and Ma, Chenglong and Chen, Jie and Zhang, Junping and Shan, Hongming},
title = {Learning to Distill Global Representation for Sparse-View CT},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2023},
pages = {21196-21207}
}
Also, welcome to visit our other works FreeSeed.
@inproceedings{ma2023freeseed,
title={FreeSeed: Frequency-band-aware and Self-guided Network for Sparse-view CT Reconstruction},
author={Ma, Chenglong and Li, Zilong and Zhang, Yi and Zhang, Junping and Shan, Hongming},
booktitle={Medical Image Computing and Computer Assisted Intervention -- MICCAI 2023},
year={2023}
}
If you have any question, please feel free to concat me at [email protected]