Hi Marc,
I assume your T2 images are undistorted.
If so, then do you have another full brain undistorted image
for each subject? If so, the best way to get the registration
to work would be to register the T2 to the full brain image
then the full brain to the diffusion image (or maybe vice versa)
then concatenate (and invert) to get the T2 into diffusion space.
Note that there is a small section on partial volume registration
in the registration practical in the FSL course:
http://www.fmrib.ox.ac.uk/fslcourse/lectures/practicals/reg/
index.html#partial
We would also recommend using fieldmaps to get the best registration
of diffusion to full brain image, via the FEAT gui (pre-stats).
If all of these images are from the same subject then it would
be better to stick with flirt for the registrations, as the
extra freedom that fnirt has will often do the wrong thing in
a within-subject registration, although to get things into
standard space (with the non-linear template) you should use
fnirt.
Hope this helps.
All the best,
Mark
On 16 Dec 2008, at 06:26, Marc Seal wrote:
> hello
>
> We have a collection of T2 data that consists of the middle 1/3 of
> the brain
> (coronal acquisition) which we want to put into diffusion weighted
> space.
>
> I was wondering if there was a FLIRT or FNIRT tweak to align an
> individual's
> partial volume to their whole brain from another acquisition?
>
> Has anyone worked out how to do this accurately and reliability?
>
> cheers,
>
> Marc
>
|