Model-Based Reconstruction for Joint Estimation of $T_{1}$ , $R_{2}^{*}$ and $B_{0}$ Field Maps Using Single-Shot Inversion-Recovery Multi-Echo Radial FLASH
This repository includes the scripts to create the Figures for the publication
Model-Based Reconstruction for Joint Estimation of
$T_{1}$ ,$R_{2}^{*}$ and$B_{0}$ Field Maps Using Single-Shot Inversion-Recovery Multi-Echo Radial FLASHX Wang*, N Scholand*, Z Tan, D Mackner, V Telezki, M Blumenthal, P Schaten, M Uecker.
Submitted to Magnetic Resonance in Medicine
Preprint on ArXiv (arXiv:2402.05366)
$^*$contributed equally to this work
This repository has been tested on Debian 12, but is assumed to work on other Linux-based operating systems, too.
Pre-processing, reconstruction and post-processing is performed with the BART toolbox.
The provided scripts are compatible with commit b9f2c33b
or later.
If you experience any compatibility problems with later BART versions please contact us!
For running the reconstructions access to a GPU is recommended.
If the CPU should be used, please remove -g
flags from all bart pics ...
, bart nufft ...
, and bart moba ...
calls.
The visualizations have been tested with Python
(version 3.9.2) and require numpy
, matplotlib
, sys
, and os
. Full list of requirements can be found in requirements.txt
. Install using pip
pip install -r requirements.txt
Ensure to have set a DISPLAY variable, when the results should be visualized.
The figures require pdflatex
, which was tested on version 3.14159265-2.6-1.40.21 (TeX Live 2020/Debian). All additionally required TeX packages can be found in installed_texlive_packages.txt
.
Install using tlmgr
tlmgr install --usermode `cat installed_texlive_packages.txt`
Additionally, the command line tool from Inkscape
1.2.2 (b0a8486541, 2022-12-01) is needed.
The data is hosted on ZENODO and must be downloaded first.
- Manual download: https://doi.org/10.5281/zenodo.10529421
- Download via script: Run the download script in the
./data
folder.- All files:
bash load-all.sh
- Individual files:
bash load.sh 10529421 <FILENAME> .
- All files:
Note: The data must be stored in the ./data
folder!
Each folder contains a README file explaining how the figure can be reproduced.
Here, the overall run time for executing ALL scripts in this repository (including the figure generation) is addressed. Individual 8 parameter IR-ME reconstructions take between 10-20 min depending on the accessible hardware.
CPU | GPU | Runtime [min] |
---|---|---|
AMD EPYC 9334 32-Core | NVIDIA H100 80G GPU | 92 |
Intel Xeon Gold 6132 | NVIDIA Tesla V100 SXM2 32 GB |
301 |
BART_GPU_GLOBAL_MEMORY=1
.
Please feel free to send us feedback about this scripts! We would be happy to learn how to improve this and future script publications.
This work is licensed under a Creative Commons Attribution 4.0 International License. You should have received a copy of the license along with this work. If not, see https://creativecommons.org/licenses/by/4.0/.