Hi,
I'm not quite sure what the FreeSurfer command is doing, so it is hard to help.
However, it might be that it makes the initial misalignment (as far as flirt is concerned) worse and that is why things are failing. If the FreeSurfer command is somehow setting the mm coordinates in the nifti qform/sform matrices to be more consistent, but changing the image contents to be more different in orientation (according to how FSLview sees them) then this would explain it. The solution then might be to add the option -usesqform in the registration and see if that helps.
All the best,
Mark
> On 22 Nov 2016, at 16:48, Dionysios Perdikis <[log in to unmask]> wrote:
>
> Dear colleagues,
>
> I am having a problem I cannot explain:
>
> I have a T1 and DWI b0 volume, that I can register (align) with flirt together successfully.
>
> However, if I use the freesurfer command mri_convert to change the orientation, size, resolution and center of the b0 volume to match the ones of T1, or the inverse, I cannot any more make the registration successfully. In fact, the output volume, e.g., the b0 volume in T1 space, has a completely wrong orientation and the resulting registration affine transform is completely wrong. Mind that in this case the resampled input b0 has the same affine transform as T1, but it is no aligned to it.
>
> Could it be a problem that I am actually asking flirt to make a registration where the input (b0 resampled to match T1), the reference (T1), and the output (b0 aligned with T1) volumes all have the same voxel to RAS affine transform?
>
> I could supply you with the relevant data and scripts if you wish.
>
> I am using the OSX development version of freesurfer:
> freesurfer-Darwin-OSX-dev-20160810-5b25567
>
> and flirt 6.0:
> FLIRT version 6.0
>
> in an iMac, Intel Core i7, OSX El Capitan 10.11.16.
>
> Thank you very much in advance for your time.
>
> Best regards,
> Dionysios Perdikis
|