MitkDiffusion

The MITK Diffusion application [1,2] offers a selection of image analysis algorithms for the processing of diffusion-weighted MR images. It encompasses the research of the Division Medical Image Computing at the German Cancer Research Center (DKFZ).

• Features & Highlights
• Image Gallery
• Downloads
• Requirements
• Building MITK Diffusion from Source
• User Manual
• References
• Contact

Screenshot of the MITK Diffusion Welcome Screen

Features & Highlights

Support for most established image formats

• Images: DICOM, NIFTI, NRRD (peak and SH images compatible with MRtrix)
• Tractograms: fib/vtk, tck and trk.

Image preprocessing

• Registration
• Head-motion correction
• Denoising
• Skull stripping and brain mask segmentation
• Resampling, cropping, flipping and merging
• Header modifications
• Single volume extraction

Diffusion gradient/b-value processing

• b-value rounding
• Gradient direction flipping
• Gradient direction subsampling
• Averaging of gradient directions/volumes
• Gradient direction and b-value visualization

ODF reconstruction and signal modelling

• Tensor and Q-ball reconstruction
• Other reconstructions via Dipy wrapping (CSD, 3D SHORE, SFM)
• ODF peak calculation
• MRtrix or camino results can be imported

Quantification of diffusion-weighted/tensor/ODF images

• Intravoxel Incoherent Motion (IVIM) and diffusion kurtosis analysis
• Calculation many other derived indices such as ADC, MD, GFA, FA, RA, AD, RD
• Image statistics

Segmentation

• Automatic white matter bundle segmentation (TractSeg) [3]
• Automatic brain mask segmentation
• Manual image segmentation and operations on segmentations
• SOON: automatic brain tissue segmentation

Fiber tractography

• Global tractography [4]
• Streamline tractography
  • Interactive (similar to [5]) or seed image based
  • Deterministic or probabilistic
  • Peak, ODF, tensor and raw dMRI based. The latter one in conjunction with machine learning based tractography [6]
  • Various possibilities for anatomical constraints.
  • Tractography priors in form of additional peak images, e.g. obtained using TractSeg

Fiber processing

• Tract dissection (parcellation or ROI based)
• Tract filtering by
  • length
  • curvature
  • direction
  • weight
  • density
• Tract resampling and compression
• Tract transformation
  • Mirroring
  • Rotating and translating
  • Registration (apply transform of previously performed image registration)
• Tract coloring
  • Curvature
  • Length
  • Weight
  • Scalar map (e.g. FA)
• Other operations
  • Join
  • Subtract
  • Copy
• Fiber clustering [7]
• Fiber fitting and weighting similar to SIFT2 and LiFE [8,9]
• Principal direction extraction (fibers --> peaks)
• Tract derived images:
  • Tract density images
  • Tract endpoint images
  • Tract envelopes

Fiberfox dMRI simulations [10]

• Multi-compartment signal modeling
• Simulation of the k-space acquisition including
  • Compartment specific relaxation effects
  • Artifacts such as noise, spikes, ghosts, aliasing, distortions, signal drift, head motion, eddy currents and Gibbs ringing
  • Definition of important acquisition parameters such as bvalues and gradient directions, TE, TR, dwell time, partial Fourier, ...
• Manual definition of fiber configurations, e.g. for evaluation purposes

Other features

• Brain network statistics and visualization (connectomics)
• Interactive Python console
• Integrated screenshot maker
• Command line tools for most functionalities

Since the last release 2017.07 there have been a lot of feature additions, bug fixes and optimizations.

Image Gallery

Scalar map visualization

Tensor Visualization

ODF visualization

Peak visualization (uniform white coloring)

Interactive tractography in MITK Diffusion. The tractogram updates automatically on parameter change and movement of the spherical seed region.

Tract dissection using manually drawn ROIs.

Result of automatic streamline weighting (similar to SIFT2 or LiFE)

Illustration of the dMRI phantom simulation process using Fiberfox.

Downloads

If you encounter any bugs, please report them in our bugtracking system or use the MITK-users mailing list. We are grateful for any feedback!

Latest stable installers (2017.07)

Commit hash 2bda849ee86a362583ee3d2beb4baaca038bd8a5

Windows 7, Windows 10	MS Windows (64 bit) installer
Windows 7, Windows 10	MS Windows (64 bit) zip archive
Ubuntu 16.04	Ubuntu (64 bit), tar.gz archive
Ubuntu 14.04	Ubuntu (64 bit), tar.gz archive

Known issues fixed in the current master: The Fiberfox command line application does not read the b-value but uses a default b-value of 1000 s/mm². This bug does not affect the GUI version of Fiberfox. This bug also has no effect if the first non-zero b-value is 1000 s/mm², which is for example the case in the simulated HCP dataset (10.5281/zenodo.572345). This bug is fixed in the current master of the MITK source code.

Requirements

• For Ubuntu users:
  • Install Python 3.X:

    sudo apt install python3 python3-pip

  • Download Python requirements file: https://phabricator.mitk.org/file/data/r7yikza26ozkodxlypsq/PHID-FILE-thn2yotzr2jein4apf3x/PythonRequirements.txt
  • Install Python requirements:

    pip3 install -r PythonRequirements.txt

  • If your are behind a proxy use

    pip3 --proxy <proxy> install -r PythonRequirements.txt

• For Windows users:
  • MITK Diffusion requires the Microsoft Visual C++ 2017 Redistributable to be installed on the system. The MITK Diffusion installer automatically installs this redistributable for you if not already present on the system, but it needs administrative privileges to do so. So to install the redistributable, run the MITK Diffusion installer as administrator.
  • Install Python 3.X: https://www.anaconda.com/download/
  • Download Python requirements https://phabricator.mitk.org/file/data/r7yikza26ozkodxlypsq/PHID-FILE-thn2yotzr2jein4apf3x/PythonRequirements.txt
  • Install Python requirementsfrom the conda command prompt:

    pip install -r PythonRequirements.txt

  • If your are behind a proxy use

    pip --proxy <proxy> install -r PythonRequirements.txt

• Requirements for all deep-learning based functionalities:
  • Affected functionalities:
    • Brain extraction
    • TractSeg
  • Pytorch: https://pytorch.org/
  • CUDA: https://developer.nvidia.com/cuda-downloads
  • (optional) cuDNN: https://developer.nvidia.com/cudnn

Building MITK Diffusion from source

1. Install Qt on your system (>= 5.11.1).
2. Clone MITK from out git repository using Git version control.
3. Configure the MITK Superbuild using CMake (>= 3.10).
   1. Choose the source code directory and an empty binary directory.
   2. Click "Configure".
   3. Set the option MITK_BUILD_CONFIGURATION to "DiffusionRelease".
   4. Click "Generate".
4. Build the project
   1. Linux: Open a console window, navigate to the build folder and type "make -j8" (optionally supply the number threads to be used for a parallel build qith -j).
   2. Windows (requires visual studio): Open the MITK Superbuild solution file and build all projects.
5. The build may take some time and should yield the binaries in "your_build_folder/MITK-build/bin"

More detailed build instructions can be found in the documentation.

References

[1] Fritzsche, Klaus H., Peter F. Neher, Ignaz Reicht, Thomas van Bruggen, Caspar Goch, Marco Reisert, Marco Nolden, et al. “MITK Diffusion Imaging.” Methods of Information in Medicine 51, no. 5 (2012): 441.

[2] Fritzsche, K., and H.-P. Meinzer. “MITK-DI A New Diffusion Imaging Component for MITK.” In Bildverarbeitung Für Die Medizin, n.d.

[3] Wasserthal, Jakob, Peter Neher, and Klaus H. Maier-Hein. “TractSeg - Fast and Accurate White Matter Tract Segmentation.” NeuroImage 183 (August 4, 2018): 239–53.

[4] Neher, P. F., B. Stieltjes, M. Reisert, I. Reicht, H.P. Meinzer, and K. Maier-Hein. “MITK Global Tractography.” In SPIE Medical Imaging: Image Processing, 2012.

[5] Chamberland, M., K. Whittingstall, D. Fortin, D. Mathieu, und M. Descoteaux. „Real-time multi-peak tractography for instantaneous connectivity display“. Front Neuroinform 8 (2014): 59. doi:10.3389/fninf.2014.00059.

[6] Neher, Peter F., Marc-Alexandre Côté, Jean-Christophe Houde, Maxime Descoteaux, and Klaus H. Maier-Hein. “Fiber Tractography Using Machine Learning.” NeuroImage. Accessed July 17, 2017. doi:10.1016/j.neuroimage.2017.07.028.

[7] Garyfallidis, Eleftherios, Matthew Brett, Marta Morgado Correia, Guy B. Williams, and Ian Nimmo-Smith. “QuickBundles, a Method for Tractography Simplification.” Frontiers in Neuroscience 6 (2012).

[8] Smith, Robert E., Jacques-Donald Tournier, Fernando Calamante, and Alan Connelly. “SIFT2: Enabling Dense Quantitative Assessment of Brain White Matter Connectivity Using Streamlines Tractography.” NeuroImage 119, no. Supplement C (October 1, 2015): 338–51.

[9] Pestilli, Franco, Jason D. Yeatman, Ariel Rokem, Kendrick N. Kay, and Brian A. Wandell. “Evaluation and Statistical Inference for Human Connectomes.” Nature Methods 11, no. 10 (October 2014): 1058–63.

[10] Neher, Peter F., Frederik B. Laun, Bram Stieltjes, and Klaus H. Maier-Hein. “Fiberfox: Facilitating the Creation of Realistic White Matter Software Phantoms.” Magnetic Resonance in Medicine 72, no. 5 (November 2014): 1460–70. doi:10.1002/mrm.25045.

Contact

If you have questions about the application or if you would like to give us feedback, don't hesitate to contact us using our mailing list or, for questions that are of no interest for the community, directly.