FastPtx: a python pTx pulse design tool for freely optimizing RF and gradient pulses with autodifferentiation

Author:

Dario Bosch <dario.bosch@tuebingen.mpg.de>

Other Contributors:

• Qi Wang <qi.wang@tuebingen.mpg.de>
  • helped with porting the bloch simulation code from C++ to Python
  • helped with pytorch support
• Alexander Loktyushin <aloktyus@tuebingen.mpg.de>
  • gave valueable input on optimization with pytorch
  • helped working out bugs

How to use

The following steps describe using Python Virtual Environments. If you prefer using Anaconda you can try adapting those steps. I don't see why it shouldn't work

1. Open a terminal in the projects' main directory
2. Create a python virtual environment: python -m venv ./env
3. Activate the environment: source ./env/bin/activate
4. Install the necessary python packages
   • pip install -r ./requirements.txt
5. Get the example data and put it into the directory ./data/ by running the downloadFiles.sh script
   • ./downloadFiles.sh
6. open either spyder or jupyter lab from the active shell
7. modify and run calc_smallFA_paper.py
   • the line dev = torch.device('cuda') sets the calculation to happen on a GPU. If you don't have a CUDA-enabled GPU, set it to cpu instead.
   • do_UP = False switches between tailored and universal pulses
   • Flip angle, pulse duration, etc can also be controlled by changing the settings in the beginning of the file.
   • Run by either executing the calc_smallFA_paper.ipynb notebook in jupyter or by running the calc_smallFA_paper.py script directly
8. You can create an animation (gif) for an optimized pulse using the animatePulse.ipynb notebook

Citing

If you use this code, please cite the corresponding paper:

Bosch, D.; Scheffler, K.: FastPtx: A Versatile Toolbox for Rapid, Joint Design of pTx RF and Gradient Pulses Using Pytorch's Autodifferentiation. Magnetic Resonance Materials in Physics, Biology and Medicine