INMR_for_DynamicMRI

Official code for "Zero-shot Implicit Neural Manifold Representation (INMR) for Ultra-high Temporal Resolution Dynamic MRI"

1. Environmental Requirements

To run the reconstruction demo, the following dependencies are required:

• Python 3.10.X (Important)
• PyTorch 2.0.0
• torchkbnufft 1.4.0
• tiny-cuda-nn 1.7
• torchvision, tensorboard, h5py, scikit-image, tqdm, numpy, scipy

2. Run the Demos

To run the basic reconstruction demo, please use the following code:

python3 main.py -g 0 -s 2 -m

To ablate the spatial C2F strategy, please use the following code:

python3 main.py -g 0 -s 2

To ablate the temporal C2F strategy, please use the following code:

python3 main.py -g 0 -s 2 -m -i 0

The rest of the parameters can be easily changed by adding arguments to the parser. The detailed definitions of the arguments can be found by:

python3 main.py -h