search

oesteban / RegSeg

RegSeg is a simultaneous segmentation and registration method that uses active contours without edges (ACWE) extracted from structural images. The contours evolve through a free-form deformation field supported by the B-spline basis to optimally map the contours onto the data in the target space.
MIT License
13 stars 9 forks source link

CSF surfaces #89

Closed oesteban closed 11 years ago

oesteban commented 11 years ago

Generate CSF surfaces for the real data example. Probably, a good idea to take a look on: https://github.com/nipy/nipype/pull/307

I also recall here an email from @meribach on April, 30th:

En VTK yo he utilizado este: http://www.itk.org/Wiki/VTK/Examples/Cxx/Medical/GenerateModelsFromLabels En ITK veo que hay esto: http://www.itk.org/pipermail/insight-users/2011-June/041770.html Ademas se que ITK SNAP lo utiliza. En Freesurfer he encontrado esto: https://surfer.nmr.mgh.harvard.edu/fswiki/mri_tessellate https://surfer.nmr.mgh.harvard.edu/fswiki/FsFastCreateSurfaceRoi Espero que te sean utiles !

oesteban commented 11 years ago

I manually merged hemispheres for white and pial surfaces with gmsh (http://geuz.org/gmsh/), but it doesn't work. The merge is not consistent, and cells are not re-computed.

I've thrown a question to the freesurfer's mailing list. We have mri_tessellate and mri_mc. The first one just puts a vertice on each corner of the on voxels. The second runs a marching cubes algorithm (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2011-November/021391.html).

oesteban commented 11 years ago

Another good ref: http://brainder.org/2012/05/08/importing-freesurfer-subcortical-structures-into-blender/

It seems sensible to use mri_pretess -> mri_tessellate -> mris_smooth

oesteban commented 11 years ago

OK, I have a pipeline to generate the CSF mesh. This is how it looks:

fa_csf

It looks much better when overlaid on the T1 (after resampling this structural in the dMRI space):

t1_csf

I upload the new code, then I will check that everything works correctly.

oesteban commented 11 years ago

Not to miss this question: https://mail.nmr.mgh.harvard.edu/pipermail/freesurfer/2013-June/030586.html