I had a look at your data, and it doesn't seem like anything is
necessarily going "wrong", but rather it seems that a lot of little
things make this a quite tricky problem for fnirt.
Your mask is quite restrictive, covering only a set of subcortical
structures. This becomes even more restrictive in the early sub-
sampling stages where a subsampled "large" voxel is included only if
_all_ the little voxels constituting the large voxel are inside the
mask. So you start out with a set of voxels that are all pretty much
sub-cortical gray matter with a very restricted range of intensities.
This means that the intensity modeling goes pretty much all over the
place for the early iterations. In particular since your two images
have quite different contrast to begin with.
I think you will be much better off if you simply don't use the mask
at all. As MJ said, because of the local displacements there isn't
really much of a case for a mask anyway. If you still want to use a
mask in order to e.g. reduce computation time you would be better off
making it more inclusive so that there is a range of intensities for
fnirt to work with. For example use a cubical ROI that encompasses all
the structures you are interested in, with some margin added on to it.
Also, you only need to specify it as the --refmask. There is no point
in including it also as an --inmask.
Good luck Jesper
On 3 Aug 2009, at 21:47, siamak wrote:
> Dear Jesper,
> Thank you so much for your time. The exact command line I am using
> is as
> fnirt --in=IBSR_09_ana_mni.nii --ref=IBSR_03_ana_mni.nii --
> 03_ana_mni_nlr_test7.nii.gz --cout=warp_coef_IBSR09-to-03_test7 --
> refmask=MNI152lin_T1_1mm_subbr_mask.nii.gz --
> --aff=IBSR09-to-03_tmp_stage1.mat --config=T1_2_MNI152_2mm.cnf
> applywarp -i IBSR_09_seg_subbr_mni.nii.gz -r IBSR_03_seg_mni.nii.gz -w
> warp_coef_IBSR09-to-03_test7.nii.gz -o IBSR09-to-
> 3dOverlap IBSR_03_seg_subbr_mni.nii.gz IBSR_03_seg_subbr_mni.nii.gz
> # Result: 132,771 , this is the original subbr volume
> 3dOverlap IBSR09-to-03_seg_subbr_mni_nlr_test7.nii.gz IBSR09-to-
> # Result: 167020, this is what we expect to be subbr of IBSR_03
> I uploaded all the files of this command with the reference number:
> Thanks for your help,
> Siamak Yousefi