A non-uniform fast Fourier transform with Kaiser-Bessel gridding for machine learning applications in PyTorch.

We are sharing a high-level PyTorch package for performing non-uniform fast Fourier transforms (NUFFTs) in machine learning. The software can be installed with a single command line on any system that already runs PyTorch.

NUFFTs are essential to reconstruction of MRI data acquired with non-Cartesian sampling trajectories. By relying on PyTorch as a back end, TorchKBNufft offers NUFFT functionality that can scale with frameworks developed by the machine learning community.

Related Publication

Muckley MJ, Stern R, Murrell T, Knoll F.
TorchKbNufft: A High-Level, Hardware-Agnostic Non-Uniform Fast Fourier Transform.
ISMRM Workshop on Data Sampling & Image Reconstruction, 2020.

Please cite this work if you are using TorchKBNufft in your research.


This resource was created and is maintained by Matthew Muckley, PhD. Questions about this resource may be raised as issues on Github.