Dear Varina,
The solution to this problem is to use fieldmaps.
I assume you do not have fieldmaps (as you have not mentioned them).
I would not use FNIRT on the functional data though, as it can easily
move things in the wrong directions (not the directions the distortions
were constrained to). Additionally, it will be highly affected by signal
dropout, leading to further incorrect warping.
I would try a 6 DOF registration from functional to highres, rather than
12 DOF, as the 12 DOF registration may introduce incorrect scalings in
order to try and compensate for the signal dropout.
If you have fieldmaps then you should definitely use them, as this is
the best way to correct distortions. We recommend that everyone
collects fieldmaps as part of their acquisition protocol.
All the best,
Mark
On 7 May 2012, at 22:38, Varina Wolf wrote:
> Hello Experts,
>
> I used linear registration with 12 DOF and full search for my functional to highres, then nonlinear for highres to standard with warp 5 mm. The warp looks perfect, but the functional to highres still has areas of func outside the highres ~ 1 - 2 cm in some areas. At the end of analysis, I think this is making activation appear outside of the brain when overlayed onto standard.
>
> Could this be improved by running FNIRT on the func to highres? If so, does the FNIRT command accept highres in place of standard? Or should I try to register from func to standard here?
>
> fnirt --ref=./highres.nii.gz --in=../denoised_data_.feat/filtered_func_data.nii.gz --iout=./func2highres_warp --subsamp=4,2,1,1 --warpres=5,5,5,5 -v -h
>
> When I try the above command from the FEAT registration folder, it did not go.
>
> Thanks for your help!
> Varina
>
|