Dear Rob,
Are you doing this outside of FEAT?
Either way, why the specification of 15 degrees?
What we normally recommend for partial field of view
imaging is the following:
1 - acquire a whole brain EPI image (just one 3D volume,
with matched resolution, TE, etc but covering the
whole brain)
2 - register the partial field of view EPI to the whole brain
EPI
3 - register the whole brain EPI to a whole brain structural
4 - register (non-linearly) the whole brain structural to the
standard space
If any of your functional or structural images are different
acquisition orientations (e.g. coronal vs axial) then it is
useful to reorient them all to the standard space orientation
first (we've recently released a tool called fslreorient2std
which does this easily for images where the nifti image is
initially labelled correctly in FSLView).
We recommend 3 DOF for step 2 above, 6 (or 7) DOF for
step 3, and a full non-linear (starting with a 12 DOF
initialisation) for step 4. We normally don't need to change
the angular search limits (and I don't recall 15 degrees
ever being tried in particular).
You also need to check each stage separately rather than
looking at the functional to standard space registration,
as this is a combination of all of the above and is subject
to a combination of all of the above. Hence you cannot
determine where any problems occur just by looking at the
functional to standard space registration result.
I hope this is helpful.
All the best,
Mark
On 18 Jul 2010, at 08:10, Rob H wrote:
> Hi
>
> We have a problem with registrating our functional data to the MNI
> standard space.
>
> Our functional data, and the anatomical image (T1 FLAIR) we are
> using are both partial brain volumes (the posterior parts of the
> brain are missing) taken in the coronal plane. On the registration
> tab we are using the T1 Flair image as our main structural image and
> the MNI 152 as the standard space.
>
> For the registration to the main structural we are using 15 degree
> search and 3DOF.
> For the registration to the standard space we are using non-linear
> transformation with 15 degree search and 7 DOF. We left the warp
> resolution as 10mm which is the default.
>
> This hasn't produced very good registration, especially between the
> functional and standard space. Is there anything anyone can suggest
> that we could do when we rerun the registration process to make it
> work better? I can post an example of the registration log tomorrow
> if that would help.
>
> Thanks
> Rob
>
|