Dear Marie,
Thanks for the images.
Unfortunately you didn't include the original TDM image (CT in English).
So I couldn't try anything myself to register the images directly.
It also seemed like this image had been processed somehow (beyond
the spatial transformation) since there was *no* signal inside the skull
whereas normally you see some brain structures, which can be helpful.
However, it appears that the big problem is non-equal scaling. The results
you got were quite reasonable in the anterior-posterior direction but were
very poor in the left-right direction. This cannot be fixed with a 6 DOF
transformation, and I suspect that the problem arises from incorrect
voxel sizes in one or more images. How did you convert to NIFTI?
Is it possible that you voxel sizes are not correct? I think that getting
the correct voxel sizes would fix your problem (and I think it is this and
not any gradient non-linearity, since it is so strong and consistent
across the different slices). If you cannot figure out what the correct voxel
size is then you could also try a 12 DOF registration, which would allow the
registration to do different scaling in different directions. It might also be
better if you can use the original image that contains more than just the
skull, as this may help the registration too.
All the best,
Mark
On 21 Mar 2011, at 12:57, Marie Eresue wrote:
> Dear Mark,
> thank you for your answer. I sent somes images on your site: number 883422.
> So I try to register "tdm_recal_spect.nii" (tdm resampled with my spect hmpao by flirt 6 dof) with my "irm.nii", by flirt 6 dof.
> I tried different cost, and I found -cost mutualinfo wich permitt have better results; I obtained this image "tdm_recal_irm5.nii" when I use flirt 6 dof . But there is a huge discrepancy on bone.
> My command is:
> flirt -dof 6 -cost mutualinfo -ref irm.nii -in tdm_recal_spect.nii -out tdm_recal_irm5.nii
>
> When I don't register spect and tdm before, I obtain image3:
> flirt -dof 6 -cost mutualinfo -ref irm.nii -in tdm_reorient.nii -out tdm_recal_irm3.nii
>
> I don't undersand why I have this problem with ct and mri .
>
> Best regards,
>
> marie
>
|