FSL

This is the official PyTorch implementation of our manuscipt:

Promoting fast MR imaging pipeline by full-stack AI
Zhiwen Wang, Bowen Li, Hui Yu, Zhongzhou Zhang, Maosong Ran, Wenjun Xia, Ziyuan Yang, Jingfeng Lu, Hu Chen, Jinfeng Lu, Jiliu Zhou, Hongming Shan, Yi Zhang
Accepted by Cell iScience

Getting started

1. Clone the repository

git clone https://github.com/wangzhiwen-scu/FSL.git
cd fsl

2. Install dependencies

Here's a summary of the key dependencies.

• python 3.7
• pytorch 1.7.1

We recommend using conda to install all of the dependencies.

conda env create -f environment.yaml

To activate the environment, run:

conda activate fsl

3. Pre-trained Model and Testing Dataset

All data and models can be downloaded in Google-drive.

It is a zip file (~843M) which contain a demo testing data and parameter files of compared models.

4. File Organization

Then place the demo testing data in:

├── datasets
│   ├── brain
│   │   ├── OASI1_MRB
│   │   ├── testing-h5py
│   │   │   ├── demo
│   │   │   │   └── oasis1_disc1_OAS1_0042_MR1.h5
│   ├── cardiac
│   └── prostate

place the parameter files in:

├── model_zoo
│   ├── pretrained_seg
│   │   └── OASI1_MRB_3seg.pth
│   └── tab1
│       └── OASI1_MRB
│           ├── asl_ablation_seqmdrecnet_bg_step3_1_local__0.05_2D.pth
│           ├── csl_seqmri_unet__0.05_2D.pth
│           ├── csmri1__0.05.pth
│           ├── csmri2__5.pth
│           └── csmtl__0.05.pth

5. Training

Please see runner/main/asl_mixed_ablation_seq_mdrec_v2_step3_1_bg_localloss.py for an example of how to train FSL.

6. Testing

bash demo.sh

Acknowledgement

Part of the subsampling learning network are adapted from LOUPE and SeqMRI. Part of the reconstruction network structures are adapted from MD-Recon-Net.

• LOUPE: https://github.com/cagladbahadir/LOUPE.
• SeqMRI: https://github.com/tianweiy/SeqMRI.
• MD-Recon-Net: https://github.com/Deep-Imaging-Group/MD-Recon-Net.

Thanks a lot for their great works!

contact

If you have any questions, please feel free to contact Wang Zhiwen {wangzhiwen_scu@163.com}.

Citation

If you find this project useful, please consider citing:

@article{wang2024promoting,
  title={Promoting fast MR imaging pipeline by full-stack AI},
  author={Wang, Zhiwen and Li, Bowen and Yu, Hui and Zhang, Zhongzhou and Ran, Maosong and Xia, Wenjun and Yang, Ziyuan and Lu, Jingfeng and Chen, Hu and Zhou, Jiliu and others},
  journal={Iscience},
  volume={27},
  number={1},
  year={2024},
  publisher={Elsevier}
}