Skip to content

PyTorch auto-differentiation based MRI pulse design, for VELOCITY SELECTIVE pulse

License

Notifications You must be signed in to change notification settings

fmrilab/VS-AutoDiffPulses

 
 

Repository files navigation

Auto-Differentiation Based MRI Pulse Design, for velocity selective pulses

Reference implementation of:
Joint Design of RF and Gradient Waveforms via Auto-Differentiation for 3D Tailored Exitation in MRI
(arXiv: https://arxiv.org/abs/2008.10594)

cite as:

@article{luo2021joint,
  author={Luo, Tianrui and Noll, Douglas C. and Fessler, Jeffrey A. and Nielsen, Jon-Fredrik},
  journal={IEEE Transactions on Medical Imaging}, 
  title={Joint Design of RF and gradient waveforms via auto-differentiation for 3D tailored excitation in MRI}, 
  year={2021},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TMI.2021.3083104}}

For the interpT feature, consider citing:

@inproceedings{luo2021MultiScale,
  title={Multi-scale Accelerated Auto-differentiable Bloch-simulation based joint design of excitation RF and gradient waveforms},
  booktitle={ISMRM},
  pages={3958},
  author={Tianrui Luo and Douglas C. Noll and Jeffrey A. Fessler and Jon-Fredrik Nielsen},
  year={2021}
}

System Requirements:

  • Ubuntu 18.04, 20.04
  • Python 3.6, 3.7, 3.8

The implementation was not tested with other configurations.

General comments

setup_AutoDiffPulses.m does the configurations for Matlab.
For the python part, in your command line, navigate to the repo's root directory, type:

pip install .

Demos are provided in ./demo.

This repo has included binary test data files for basic accessibility in certain regions.
Future binary data files will be added to: https://drive.google.com/drive/folders/1EyKhA_d74OC4KADMuTd1kRTEMoVqWdIY.

Dependencies

This work requries Python (≥v3.5), PyTorch (≥v1.3) with CUDA.

  • MRphy: Python, Github link (≥v0.1.8).
  • +mrphy: Matlab, Github link.
  • +attr: Matlab, Github link.

Other Python dependencies include:
scipy, numpy, PyTorch.

About

PyTorch auto-differentiation based MRI pulse design, for VELOCITY SELECTIVE pulse

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 79.5%
  • MATLAB 20.5%