LMPDnet

Deep unrolled primal dual network for TOF-PET list-mode image reconstruction

Set up

Create a conda Environment

conda env create --name LMRecon python==3.8
conda activate LMRecon
pip install requirements.txt

Set system environment variables

1. Download the file parallelproj.c_dll and create a new variable named PARALLELPROJ-C-LIB in the system variable, pointing to the path where parallelproj_c.dll is located
2. Download the file parallelproj_cuda.dll and create a new variable named PARALLELPROJ-CUDA-LIB in the system variable, pointing to the path where parallelproj_cuda.dll is located
3. Create a new variable named PYTHONPATH in the system variable to point to the current working directory.

Acknowledgement

Our simulation and projection codes are based on parallelproj. The simulated phantom is created based on the FBSEM, Many Thanks.

Citation

If you find our paper or repo useful, please consider citing our paper:

@article{Hu_2025,
    doi = {10.1088/1361-6560/adf9b7},
    url = {https://doi.org/10.1088/1361-6560/adf9b7},
    year = {2025},
    month = {aug},
    publisher = {IOP Publishing},
    volume = {70},
    number = {17},
    pages = {175012},
    author = {Hu, Rui and Li, Chenxu and Tian, Kun and Cui, Jianan and Chen, Yunmei and Liu, Huafeng},
    title = {Deep unrolled primal dual network for TOF-PET list-mode image reconstruction},
    journal = {Physics in Medicine & Biology},
}