Projects/Slicer3/2007 Project Week Slicer Matlab Pipeline for scalars and tensors

From NAMIC Wiki
Revision as of 03:02, 29 June 2007 by Kquintus (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < Projects < Slicer3 < 2007 Project Week Slicer Matlab Pipeline for scalars and tensors
Slicer Matlab Pipeline Schema

Key Investigators

  • BWH: Katharina Quintus
  • Isomics: Steve Pieper
  • BWH: Sylvain Bouix
  • BWH: Marc Niethammer


  • Provide tools to pipe volume data from Slicer3 memory to Matlab and back.

Approach, Plan

  • Finalize utilities that transform tensor data from Slicer's IJK space to the original gradient space and transform the data from gradient space back to Slicer's IJK space respectively.
  • Quantify numerical error for these transformations.
  • Application testing: Get user feedback.


  • Pipeline tools for scalar data are ready to use
  • adapted Slicer Matlab interface to recent changes in vtkNRRDReader
  • added Matlab functions:
    • load nrrd volume from file into a Matlab structure that also holds header information
    • save Matlab nrrd structure to file
    • transform tensor data from Slicer IJK space into diffusion gradient space and vice versa
  • to do:
    • replicate these functions in Teem library
    • quantify error for IJK to gradient space transformation


Slicer Deamon Wiki page

Additional Information

When Slicer is started with the Slicer Daemon enabled (either with the "--daemon" flag or by turning on the Daemon using he View->Application Settings dialog's Slicer Options tab), a server socket is created that is listening and waiting for new connections. This network service can be used to access the MRML scene or other objects in Slicer memory. Several clients have been written or are beeing worked on at the moment:

  • Tcl scripts that read out volumes to stdout or write to stdin
  • Python based clients
  • Matlab clients

The Slicer Deamon Wiki page provides more details.

The Matlab client

  • Matlab scripts: getSlicerVolume.m and putSlicerVolume.m use Matlab extention popen to connect to stdout(stdin respectively) of the tcl client slicerget.tcl (slicerput.tcl respectively). The tcl client establishes a channel to the SlicerDaemon socket and requests(sends) data.
  • flag in these scripts can be set to do/undo gradientSpace->IJKspace transformation for tensor data.


The Matlab pipeline client can be used to send data from Slicer to Matlab, where data can be processed making advantage of already existing code and powerful Matlab toolboxes. The data can then be sent back to Slicer for visualisation or further processing. No file I/O is necessary. This can be very useful since a lot of methods or algorithms are first prototyped in Matlab. Ideas what can be done easily in Matlab while Slicer does not have the functionality yet:

  • Tensors:
    • Thresholding
    • Do tensor statistics after masking tensor with labelmap
    • Average tensors
    • Smooth tensor fields
  • Scalars:
    • plot histogram (that allows to quantify, at the moment slicer-histograms are not associated with numbers )

Tensor transformation issues

When tensors are loaded into Slicer, the vtkNRRDReader performs transformation from diffusion gradient space into image space (IJK space). This is done to improve performance since most itk-filters require data in ijk coordinates.

Tensors piped to Matlab by default will live in IJK-space, since this is where they are in Slicer memory. Assuming the dataset will be sent back to Slicer after Matlab processing, no such transformation needs to be done. Besides, this transformation and especially the inverse transformation when putting data back into Slicer will introduce numerical errors due to matrix inversion.

However, the Matlab user might wish to transform the data back to gradient space. Matlab tools are provided to perform transformations needed.

Issues with different coordinate conventions in Matlab versus Slicer

The pipeline user in Matlab has to be aware of the fact that when the volume is visualized in Matlab, the image will appear to be transposed due to different coordiate conventions.

Back to Project week