Toolbox for Computational Magnetic Resonance Imaging

The Berkeley Advanced Reconstruction Toolbox (BART) toolbox is a free and open-source image-reconstruction framework for Computational Magnetic Resonance Imaging developed by the research groups of Martin Uecker (Göttingen University) and Michael Lustig (UC Berkeley). It consists of a programming library and a toolbox of command-line programs. The library provides common operations on multi-dimensional arrays, Fourier and wavelet transforms, as well as generic implementations of iterative optimization algorithms. The command-line tools provide direct access to basic operations on multi-dimensional arrays as well as efficient implementations of many calibration and reconstruction algorithms for parallel imaging and compressed sensing.

mri scan of human abdomen

Figure: l1-ESPIRiT reconstruction of a human abdomen (variable-density Poisson-disc sampling, R=7, RF-spoiled 3D-FLASH, B0 = 3T TR/TE = 4.3/1.0ms, partial echo .6, matrix: 320x256x184, 32 channels)
Total reconstruction time: 51s including compression (14s), calibration (9s), iterative reconstruction (12s), and other processing steps (16s) on a multi-GPU system.

List of Features

  • basic features:
    • support for Linux, Mac OS X, and Windows (with Cygwin)
    • multi-dimensional operations on arrays
    • fast non-uniform Fourier Transform (nuFFT)
    • multi-dimensional (divergence-free) wavelet transform
    • parallel computation on multiple cores and with Graphical Processing Units (GPU)
  • iterative methods:
    • Conjugate Gradients (CG)
    • (Fast) Iterative Soft-Thresholding Algorithm (ISTA and FISTA)
    • Alternating Direction Method of Multipliers (ADMM)
    • Iteratively Regularized Gauss-Newton Method (IRGNM)
  • calibration methods for parallel MRI:
    • direct calibration from k-space center
    • Walsh's method
    • ESPIRiT
  • reconstruction methods for MRI:
    • iterative parallel imaging reconstruction: POCSENSE, SENSE
    • compressed sensing and parallel imaging
    • non-linear inverse reconstruction: NLINV (blind multi-channel deconvolution)
    • calibration-less parallel imaging: SAKE (structured low-rank matrix completion)
  • regularization (in arbitrary dimensions):
    • Tikhonov
    • total variation
    • l1-wavelet
    • (multi-scale) low-rank

Example

Perform ESPIRiT calibration and image reconstruction with l1-wavelet regularization:

 $ bart ecalib kspace sensitivities
 $ bart pics -l1 -r0.001 kspace sensitivities image_out

A python-based image viewer (bartview.py) which can read the BART data format is included in the source repository.

An image viewer for Linux and Mac OS X which is currently in development can be found here.

Mailing List

Please direct all questions or comments to the public mailing list:

mrirecon list (public)

(or contact the main author: martin.uecker at med.uni-goettingen.de)

Download

Note: The software is intended for research use only and NOT FOR DIAGNOSTIC USE. It comes without any warranty (see LICENSE for details).

Releases:

It is recommended to download the latest release. The latest release can always be found here.
Attention: Windows users should still use version v0.3.01 because a library is missing on cygwin which is requited to compile later versions.

BART has also been included in Debian GNU/Linux. The Debian binary package can be reproducibly built from the source code (as distributed by Debian) and can be downloaded from here. The package should also work on Ubuntu although this is not guaranteed.

For developers: the C source code can be found in the GitHub repository

Installation

Installation of the required libraries, downloading and unpacking of the archive, and compilation on Linux is usually as simple as typing the following commands:

 $ sudo apt-get install make gcc libfftw3-dev liblapacke-dev libpng-dev libopenblas-dev
 $ wget https://github.com/mrirecon/bart/archive/vX.Y.ZZ.tar.gz
 $ tar xzvf vX.YY.ZZ.tar.gz
 $ cd bart-X.YY.ZZ
 $ make

See the README file included with the source code for further instructions and for Mac OS X and Windows.

If you are a Docker user you can also start with this extremely simple Dockerfile.

Matlab Interface and Examples

The toolbox can also be used in combination with Matlab/Octave.

 >> sensitivities = bart('ecalib', kspace);
 >> image_out = bart('pics -l1 -r0.001', kspace, sensitivities);

More examples where the tools are called directly from Matlab can be found here.

Matlab code and data: GitHub repository

A Matlab-based image viewer which works well with BART is arrayShow by Tilman Sumpf.

Workshop Materials with Examples

The toolbox was presented at the ISMRM 2016 Data Sampling and Image Reconstruction Workshop. This material was created for BART version 0.3.00 and later versions might have minor differences. Please check the README included with each release for up-to-date installation instructions.

Demo code and data: GitHub repository

References

A a generic reference (all versions): BART Toolbox for Computational Magnetic Resonance Imaging, DOI: 10.5281/zenodo.592960