Salt Lake AHM:DTI Workgroup TCON

Attendees: Bill, Ross, Gordon, Guido, Martin, Isabelle, Tina, Lauren, Dave Tuch, Jim Miller, C-F Westin, Will, Luis, Andy, Steve, Eric

[Ross]

• need to talk about representation of DTI data and mechanisms for file IO

[Guido/Martin/Isabelle]

• support for tractography: representing fiber tracts
• how to represent tensor data in ITK (proposal in dti community pages on Wiki). Probably need a special data type, perhaps like a tensor type in ITK (similar to pixel).
• how to support non-tensor data, such as QBALL
• need to talk about display of DTI data also
• q: should a discussion of spatialobjects be included? a. [Steve] group should be brought upto date on the work that Julien has been doing in the last few days about spatialobjects and slicer. [Bill]: agree.

[Gordon]

• NRRD format for representing tensor data: stores mask alongside the tensor values. This is a 4d array which holds 7 values, one for mask and 6 for tensor. (This is supported in ITK.) [GK: this is a convention for using NRRD to store diffusion tensors; NRRD can store N-D arrays]
• general orientation support has been added to NRRD. how to map this to internal representation of ITK is TBD.
• will be attending SLC AHM.
• will add NRRD information to DTI community web page.
• NIFTI extension to analyze can store orientation but not mask info.
• only argument about file formats should be how expressive they are.
• formats Dave Tuch referred to: MINC http://www.bic.mni.mcgill.ca/software/minc/ and NIFTI http://nifti.nimh.nih.gov/

[Dave Tuch]

• q. Support for standard DTI formats? a. (Bill?) I/O Factory
• VTK/ITK support for irregularly spaced tensors, eg, for cortical surface mapping.
• Perspective from ISMRM Diffusion & Perfusion MR Study Group
• Tensors in general (eg, deformation tensors) or just diffusion tensors?

[Bill]

• IO factory mechanism in ITK supports many popular formats. The apps don't need to be aware of formats. Turned out pretty nice.
• ITK is on it's 3rd DICOM implementation -- the process we will advocate for NA-MIC is to start with a set of requirements which will evolve over time. Might seem chaotic at the beginning, but has worked out well for the ITK community. Data structures that will be picked to represent DTI will need to be more stable though.
• Get a list of use-cases for how people are using DTI data.
• We may want to hold a similar TCON for fMRI, which is a more mature modality.

[Will]

• Q: are DTI tensors really symmetric? answer [Gordon]: physics dictates that they are.

[Steve]

• Should think involving core 3 people in the discussion at some point.