Intersubject

From NAMIC Wiki
Revision as of 13:21, 18 December 2006 by Andy (talk | contribs) (Update from Wiki)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < Intersubject

Goal

The data set below should be used for testing non-rigid registration methods targeted towards intersubject mapping of anatomical MRIs, specifically cortical regions (MR morphology to MR morphology). Please pay especially attention to voxels within areas of similar intensities as "standard" methods, such as Mexwell Demons or B-Spline, randomly map them from source to target.

Requirements

Any solution posted has to fullfill the following requirements:

  • relatively robust, with few parameters to tweak
  • runs on grey scale images
  • has already been published
  • relatively fast (ideally speaking a few minutes for volume to volume).
  • not patented
  • can be implemented in ITK and parallelized.

If your method fullfills these requirements please modify the [#Test_Script test script], run it on the [#Data data set] , and [#Post_Results post your results].

Test Script

To test the accuracy of your approach please modify the following tcl script and donwload the newest Slicer Vesion 2.7. If you think the script is to complicated just make sure that you

  1. Register all cases to a CASEXX using your method
  2. Apply your resampling method (only use nearest neighbor !) to the labelmaps and corresponding deformation fields
  3. Measure Dice Score between segmentation of CASEXX and deformed segmentation.

Data

data and the segmentations.

If you cannot access the data please email Kilian Pohl.

Posted Results

Please post your results

  • Download any files to this wiki that you link to in your description. The exceptions to the rule are software and source code.
  • Post your results in the following the template (edit this page to view template)

Posted results:

  1. Kilian Pohl (BWH): B-Spline implementation by Rohlfing
    • Description: Our implementation of nonrigid registration is a modified version of an algorithm first described by Rueckert et al. It also incorporates some features of methods presented by other groups. The geometric transformation model is based on free-form deformations represented by multilevel B-splines [29], defined on a uniform three-dimensional (3-D) control point grid (CPG). The control points are moved independently and define a continuous deformation of the coordinate space by interpolation between them using 3-D third-order B-splines.
      T. Rohlfing and C. R. Maurer, Jr., “ Nonrigid image registration in shared-memory multiprocessor environments with application to brains, breasts, and bees,” IEEE Transactions on Information Technology in Biomedicine, vol. 7, no. 1, pp. 16-25, 2003.
    • Availability of Software: Not available
    • Test Result:
      • Overal Dice Score: % , [[media: | Test-File]]
      • Warped images: [[media: | tgz file of warped images]]
      • Analysis: Describe Test results. If possible, compare to other postings.