MRI super-resolution reconstruction using efficient diffusion probabilistic model with residual shifting

🔥🔥Res-SRDiff is a deep learning framework designed to robustly restore high-resolution pelvic T2w MRI and ultra-high field brain T1 maps using an efficient probabilistic diffusion model.

• Our paper on arXiv: MRI super-resolution reconstruction using efficient diffusion probabilistic model with residual shifting ❤️

🔍 Diffusion Process

The following diagram illustrates the diffusion process used in this project:

Getting Started

⚠️ Important: The code for this study will be released publicly upon acceptance of the manuscript .

Prerequisites

• Python (>=3.12)
• PyTorch (>=2.5)
• NVIDIA CUDA (for GPU acceleration)
• Additional dependencies as listed in requirements.txt

Installation

1. Clone the repository:

   git clone https://github.com/mosaf/Res-SRDiff.git
   cd Res-SRDiff
   

2. Install dependencies:

   conda env update --file environment.yml --prune
   

Running the Code

To run the project, modify the parameters in the main.py file and execute the main.py script:

python main.py

⚙️ Model Hyper-parameters

The diagram below visualizes the key hyper-parameters used in this model:

📚 Citation

If you find Res-SRDiff useful for your research or project, please consider citing our work:

@misc{safari2025mrisuperresolutionreconstructionusing,
      title={MRI super-resolution reconstruction using efficient diffusion probabilistic model with residual shifting}, 
      author={Mojtaba Safari and Shansong Wang and Zach Eidex and Qiang Li and Erik H. Middlebrooks and David S. Yu and Xiaofeng Yang},
      year={2025},
      eprint={2503.01576},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.01576}, 
}

Acknowledgments

• This project is based on Original Repository Name.