From NITRC Wiki

References

• B. A. Landman, H. Wan, J. Bogovic, P.-L. Bazin, and J. L. Prince. “Resolution of Crossing Fibers with Constrained Compressed Sensing using Traditional Diffusion Tensor MRI”, In Proceedings of the SPIE Medical Imaging Conference. San Diego, CA, February 2010
• B. A. Landman, H. Wan, J. A. Bogovic, P. C. van Zijl, P-L. Bazin, and J. L. Prince. “In the Pursuit of Intra-Voxel Fiber Orientations: Comparison of Compressed Sensing DTI and Q-Ball MRI”, International Society for Magnetic Resonance in Medicine, Stockholm, Sweeden, May 2010
• B. A. Landman, H. Wan, J. A. Bogovic, P. C. van Zijl, P-L. Bazin, and J. L. Prince. “Accelerated Compressed Sensing of Diffusion-Inferred Intra-Voxel Structure through Adaptive Refinement”, International Society for Magnetic Resonance in Medicine, Stockholm, Sweeden, May 2010
• B. A. Landman, J. Bogovic, and J. L. Prince. “Compressed Sensing of Multiple Intra-Voxel Orientations with Traditional DTI”, In Proceedings of the Workshop on Computational Diffusion MRI at the 11th International Conference on Medical Image Computing and Computer Assisted Intervention, New York, NY, September 2008.
• More details: https://masi.vuse.vanderbilt.edu/index.php/MASI:Publications

CFARI

Diffusion tensor imaging (DTI) is widely used to character- ize tissue microarchitecture and brain connectivity. However, traditional tensor techniques cannot represent multiple, independent intra-voxel ori- entations, so DTI suffers serious limitations in regions of crossing fibers. We present a new application of compressed sensing, Crossing Fiber Angular Resolution of Intra-voxel structure (CFARI), to resolve multiple tissue orientations. CFARI identifies a parsimonious tissue model on a strictly voxelwise basis using traditional DTI data. Reliable estimates of multiple intra-voxel orientations are demonstrated in simulations, and intra-voxel fiber orientations consistent with crossing fiber anatomy are revealed with typical in vivo DTI data.

CFARI has always been, is, and will be available in open source through the JIST project. I (Bennett) am currently (June 2010) preparing our first journal publication on this line of research. We are also exploring ways that we might may good use of the multi-directional structure estimated from low-b-value diffusion imaging. The software should be considered "beta". Currently, the layouts have a bit of inefficiencies due to required reformatting of the 4-D volumes for various modules. Please report any problems or feature change requests to the forums.

User Manual

1. Get JIST and MIPAV

• This tutorial is written for JIST Beta 1.10+ using MIPAV 4.4.1-NIGHTLY-June-3-2010+. More recent versions of both packages should work. Please report any problems. Instructions: jist:MainPage#Getting_Started

2. Get Some Data

• We have used CFARI on many different types of imaging data - DTI, q-ball, generic HARDI, simulated, empirical, etc. For this tutorial, we'll use a dataset from the Multi-Modal Reproducibility Resource: http://www.nitrc.org/projects/multimodal/
• Get our project (zip file with data, layout, misc files): http://jist.projects.nitrc.org/CFARIdemo.tar.gz
• Unzip the zip file. Locate the following text files:
  • dti.txt
  • grad.txt
  • b.txt
• Change the path in each of these text files to point to the location on your local system where you unzipped the file.
• Save the text files.

3. Configure the Layout In Jist

• Start Mipav and start the JIST layout.
• Change your GlobalPreferences to use .LayoutXML (the new format) rather than .Layout (the old format)
• Optionally, change your global preferences to use a different data format -- I use NIFTI (NII) with GZIP compression.
• Open the included CFARI.LayoutXML file.
• Change the Layout preference so that:
  • the data are written to a directory of your choice
  • the processes use a reasonable amount of RAM for your system
  • the processes use a reasonable number of CPU's for your system
• In the main layout window, select each of the three purple input modules. After selecting each one,
  • Change the source file ("List File") to point to the file of the same name in the location where you unzipped the demo.
• Select the "Define Camino Scheme" module. Update the last three inputs (bigdelta.txt, smalldelta.txt, te.txt) to point to the files of the same name in your unzipped project directory.

4. Run the Layout

• From the Layout Tool, select Project->Process Manager
• Several task status lines should read "READY" while most should read "NOT READY"
• Select Scheduler->Start Scheduler
• Wait ~ minutes.
• After awhile, tasks should change from NOT READY->READY->RUNNING->COMPLETED.
• If tasks show "FAILED", then right click on the task to get debugging information. These debugging reports are also text files in the corresponding directory. Please include debug reports (both output and error) with bug reports to the forum.

5. Analysis

• The directions are reported in two files - a set of indexes into the directional basis set (the larger of the two basis sets) and the set of mixture fractions.
• No automated tools have yet been published for examining these directions. We have used:
  • Matlab and Paraview to visualize these datastructure
  • Matlab and custom fiber tracking code to perform analyses
  • Matlab to assess reproducibility
• Feedback is welcome as to how to make directional reports more useful.

6. Options

• You can adjust the orientation of reconstruction basis set by modifying the options to "inscribe platonic solid" and "tessellate surface"
• You can adjust the canonical tensor of the basis set in the "Define CFARI Scheme" module.
• By default, only the top 5 directions reported. This can be changes in the CFARIEstimation module.