AHM2010:DiffusionDatatypesBreakout

Agenda

• How should the NAMIC-Kit handle new upcoming diffusion datatypes?
• Discussion of
  • Varieties of diffusion datatypes
  • Slicer data model for visualization of new datatypes
  • File formats for interoperability of analysis tools

Overall Conclusions

• All attendees want to use Slicer3 for visualization in future. They agree Slicer3 generic data model should enable plugins of new data for diffusion visualization. This will be very useful.
• All attendees agree vtkPolyData for fibers, and nrrd for volumes, are sufficiently generic to store all datatypes we can imagine.
• See below for meeting notes.

Logistics:

• Wednesday January 6, 12:30 pm - 1:30 pm
• Location: Amethyst 1

Attendees

• Anuja Sharma
• Alex Yarmarkovitch
• Mahnaz Maddah
• Deepika Mahalingam
• Yundi Shi
• Zohara Cohen
• Martin Styner
• Gopal Veni
• James Malcolm
• Luke Bloy
• Lauren O'Donnell

Meeting Notes

DATA TYPES LIST (not exhaustive, includes data used by attendees and other well-known models)

• real spherical harmonic coefficients: vectors but of varying lengths 6 to 100 components, tied to a basis
• two-tensor or mixtures of tensors, of watson functions
• mixture models in general
• PAS vector of coefficients, basis
• Gaussian mixture models - Fisher basis gaussian on sphere
• ball and stick model (Tim Behrens)
• probabilistic tracking -- scalar map
• DSI - volumes full FT of signal
• qball
• (various scalar measures from the above)

FRAMEWORK IDEAS (to enable development of slicer visualization for new data)

• Visualization issue: difficult to implement in current framework (Luke Bloy experience involved duplicating all code from tensors for spherical harmonics).
• Everyone would like to use slicer to look at their data in future
• Data Model Class Abstraction Idea
  • overloaded methods/ plugin idea
  • API for new data - how to view it in slicer
    • implement the following methods i.e. display node, scalar computation
    • at this point give me a glyph
    • at this point give me scalars that are relevant
  • Use polydata and add arrays to it for multicomponent data
  • Each data model class knows how to deal with its data
  • Model should ideally be used for volumes/models/fibers

INTEROPERABILITY (file types for cross-program diffusion data I/O)

• current file type is vtkPolyData - named arrays can tag datatypes for future datatype expansion.
• nrrd for volumes - for future, allows arbitrary strings/fields if we need to store metadata
• Conclusion is these formats are sufficiently general and can remain the standard for NAMIC interoperability

Back to NA-MIC AHM 2010