xmed-lab / DIF-Net

MICCAI 2023: Learning Deep Intensity Field for Extremely Sparse-View CBCT Reconstruction
MIT License
60 stars 11 forks source link
implicit-neural-representation miccai miccai-2023 nerf reconstruction

DIF-Net

Yiqun Lin, Zhongjin Luo, Wei Zhao, Xiaomeng Li, "Learning Deep Intensity Field for Extremely Sparse-View CBCT Reconstruction," MICCAI 2023. [paper]

0. Citation

@inproceedings{lin2023learning,
  title="Learning Deep Intensity Field for Extremely Sparse-View CBCT Reconstruction",
  author="Lin, Yiqun and Luo, Zhongjin and Zhao, Wei and Li, Xiaomeng",
  booktitle="Medical Image Computing and Computer Assisted Intervention -- MICCAI 2023",
  pages="13--23",
  year="2023",
  publisher="Springer Nature Switzerland"
}

1. Installation

torch 1.8.0
numpy, opencv-python, SimpleITK

2. Data Preparation

Please follow the scripts (4 steps) given in ./data/knee_cbct/*.npy to conduct preprocessing. For detailed instructions, please refer to ./data/knee_cbct/README.md. The processed data will be organized as follows.

├── ./data/knee_cbct/
│   ├── config.yaml
│   ├── info.json
│   ├── processed/
│   │   └── FL-140400.nii.gz
│   ├── blocks/
│   │   ├── blocks.npy
│   │   ├── FL-140400/
│   │   │   ├── block_0.npy
│   │   │   ├── block_1.npy
│   │   │   └── ...
│   ├── projections_normalized/
│   │   └── FL-140400.pickle

3. Training and Testing

Follow the scripts given in ./scripts/*.sh to conduct training and testing.

License

This repository is released under MIT License (see LICENSE file for details).