GUST-Flow

Gaussian-based unsupervised phase unwrapping for 4D Flow MRI.

Installation

git clone https://github.com/AssociatedPrimeIdeal/GUST-Flow.git
cd GUST-Flow
pip install .

A CUDA-enabled PyTorch installation and a compatible NVIDIA CUDA toolkit are required. The Python package is gustflow.

Reproducible Demo

Run test.ipynb from the repository root. It loads TestData.h5, reconstructs velocity using GUSTFlow, and exports a training GIF.

TestData.h5 is a demonstration dataset derived from FlowVN, cropped along the temporal dimension to remain below 100 MB.

License

See LICENSE.