PGIUN:Physics-Guided Implicit Unrolling Network for Accelerated MRI

This repository is the official implementation of PGIUN:Physics-Guided Implicit Unrolling Network for Accelerated MRI, accepted by TCI. If you have any questions, please feel free to contact me："[email protected]"

Requirements

To install requirements:

pip install -r requirements.txt

einops==0.4.1
ipdb==0.13.9
layers==0.1.5
matplotlib==3.5.2
numpy==1.20.3
PyYAML==6.0.1
scikit_image==0.19.3
scipy==1.7.3
setuptools==59.5.0
SimpleITK==2.3.1
thop==0.1.1.post2209072238
timm==0.5.4
torch==1.13.1+cu117
torchvision==0.14.1+cu117
tqdm==4.64.1

Dataset Setup

Data
│   ├── T1
│   │   ├── train
│   │   │   ├── train_1.npy         
│   │   │   ├── train_2.npy 
│   │   │   ├── ...         
│   │   │   └── train_N.npy
│   │   └── valid
│   │   │   ├── valid_1.npy         
│   │   │   ├── valid_2.npy 
│   │   │   ├── ...         
│   │   │   └── valid_N.npy
│   │   └── test
│   │   │   ├── test_1.npy         
│   │   │   ├── test_2.npy 
│   │   │   ├── ...         
│   │   │   └── test_N.npy
│   │   
│   ├── T2
│   │   ├── train
│   │   │   ├── train_1.npy         
│   │   │   ├── train_2.npy 
│   │   │   ├── ...         
│   │   │   └── train_N.npy
│   │   └── valid
│   │   │   ├── valid_1.npy         
│   │   │   ├── valid_2.npy 
│   │   │   ├── ...         
│   │   │   └── valid_N.npy
│   │   └── test
│   │   │   ├── test_1.npy         
│   │   │   ├── test_2.npy 
│   │   │   ├── ...         
│   │   │   └── test_N.npy
│   │   
│   │   └── ...
│   └── ...
│            
└── ...

Configure data_dir and root_path in the config.yaml folder, and configure the config.yaml path in option.py.

Training

To train the model(s) in the paper, run this command:

python mc_rec_main.py --model pgiun --batch_size 1 --n_epochs 100 --mask random --gpuid 0 --modal T2 --acceleration 4 --data_name IXI

where
--model provides the model name for the current run.
--mask provides the mask used in the current run.
--acceleration defines the acceleration ratio.
--data_name provides the data name of the current run.
Other hyperparameters can be adjusted in the code as well.

Evaluation

To evaluate the model on MRI dataset, e.g., IXI, BraTS, fastMRI, run:

python mc_rec_main.py --model pgiun --batch_size 1 --n_epochs 100 --mask random --gpuid 0 --modal T2 --acceleration 4 --data_name IXI --train test

Citation

If you find it helpful, please cite our literature:

@article{jiang2024pgiun,
  title={PGIUN: Physics-Guided Implicit Unrolling Network for Accelerated MRI},
  author={Jiang, Jiawei and He, Zihan and Quan, Yueqian and Wu, Jie and Zheng, Jianwei},
  journal={IEEE Transactions on Computational Imaging},
  year={2024},
  publisher={IEEE}
}