[ITK] Voxel classication Kmeans and MRF

[Cyril Jaudet]

Hello itk community,

I am trying to classify some PET images in different populations. The goal
is to separate the voxel in N population if they differences is superior to
the inaccuracy of the acquisition.

For that purpose i use first a k means classifier with a number of class
egal to max value- min value divide by standard deviation. It seems to work
well.

The second step will be to use the MRF classifier to account of the spatial
distribution of the voxels.

I have two problems to begin.

1) I don't know how to use the label map created by kmeans  as a starting
point. In the Insight software guide the labelmap is read but not use, i
think?

2) In the insight guide  if SetNeighborhoodRadius( 1 ) in a 3d image the
number of Neighbor isn't 8+1 (it's in 2d?) but 26+1 central?

Thank's you for your time.
Best regards,

Cyril Jaudet
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/community/attachments/20150505/0e07baa7/attachment.html>