Rapid Spatio-temporal MR Fingerprinting Using Physics-Informed Implicit Neural Representation

This repository provides the official implementation of the paper: Rapid Spatio-temporal MR Fingerprinting Using Physics-Informed Implicit Neural Representation (paper link)

• Introduction
• Project Structure
• Getting Started
• Run
• Citation
• Contacts

Introduction

πMRF (Physics-informed implicit neural MRF) is a physics-informed unsupervised framework for accurate quantitative parameter mapping via global spatio-temporal inversion.

Project Structure

The main components of this repository are organized as follows:

piMRF/
├── main.py              # Runnable demo script for running πMRF reconstruction.
├── model/               # Core implementation of networks and solvers.
│   ├── model.py         # Network architectures (NablaBlochNet, DinerSiren) and loss functions.
│   └── piMRF.py         # πMRF reconstruction solver implementation.
├── configs/             # Configuration files.
│   ├── config.json      # Parameter settings for the reconstruction.
│   └── piMRF_config.py  # Configuration loading and merging logic.
├── utils/               # Utility functions and operators.
│   ├── utils.py         # Data processing and visualization utilities.
│   ├── SIM_EPG.py       # Bloch simulation (EPG) implementation.
│   └── Nufft_multi.py   # Multi-coil NUFFT operators.
├── data/                # Example datasets including k-space, trajectories, and maps.
└── results/             # Directory for saving reconstruction outputs and logs.

Getting Started

The hardware and software environment we tested:

• OS: Ubuntu 22.04.5 LTS
• CPU: Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz
• GPU: NVIDIA A40 48GB
• CUDA: 13.0
• PyTorch: 2.7.1
• Python: 3.11.13

Installation

0. Download and Install the appropriate version of NVIDIA driver and CUDA for your GPU.
1. Download and install Anaconda or Miniconda.
2. Clone this repo and cd to the project path.

git clone https://github.com/Ashgon/piMRF.git
cd piMRF

3. Create and activate the Conda environment:

conda create --name piMRF python=3.11.13
conda activate piMRF

4. Install dependencies:

pip install -r requirements.txt

Run

To run the reconstruction demo, please use the following command:

python main.py

Reconstruction results are written to the results/ folder.

Citation

If you find this code useful, please cite our work:

@article{RapidSpatiotemporalMRgong2026,
  title = {Rapid Spatio-Temporal MR Fingerprinting Using Physics-Informed Implicit Neural Representation},
  author = {Gong, Chaoguang and Zou, Lixian and Li, Peng and Wu, Xingyang and Qiao, Yangzi and Hu, Zhanqi and Wang, Xiaoyan and Zhou, Yihang and Wang, Kai and Hu, Yue and Wang, Haifeng},
  year = 2026,
  month = mar,
  journal = {Medical Image Analysis},
  volume = {109},
  pages = {103935},
  doi = {10.1016/j.media.2026.103935}}

Contacts

• Chaoguang Gong (23b305002@stu.hit.edu.cn)
• Haifeng Wang (hf.wang1@siat.ac.cn)
• Yue Hu (huyue@hit.edu.cn)