Hi,
It might help, if you orient your epi volumes to roughly match the
anatomical. Than cut down the anatomical scan that more or less that
part of the brain is left that is covered by your epi slices (either way
should work, fslmaths with the -roi option, or the fslroi function,
though i prefer the second). With this reduced FOV of the anatomical
data you might get better results. If you need to transform the epi scan
in several steps, you might want to combine the transformation matrices
with convert_xfm and apply the new one to the initial data set to reduce
the amount of interpolation.
Good luck,
wolf
On 02/15/2011 06:19 PM, Cris Lanting wrote:
> Dear all,
> I am aware that this issue comes around every now and then on the list
> but I haven't found a work-around yet...
> I have some difficulties with registering 18 functional slices (1mm3
> isotropic) to a full brain MPRage anatomy file (also 1mm3 isotropic).
> I tried all the obvious methods (i.e. playing around with the degrees
> of freedom; use the 'xyztrans' and 'ztransonly' schemes) but no luck
> so far.
>
> Although I now know that things may improve by acquiring a full-brain
> EPI image I haven't got one for the initial data sets but there is one
> option that works: Thankfully, the orientation info from the scanner
> is very good at aligning the epi and
> mprage and what I do is the following:
>
> flirt -in epi.nii.gz -ref mprage.nii.gz -usesqform -applyxfm -out
> func2mprage -omat func2mprage.mat
>
> This creates a good registration!
>
> What I now want is the apply the same transformation in Feat where I
> try to register the data to the main structural image (mprage) which
> in turn is registered to standard space. Yet seems that I only can use
> the normal methods that, in turn, fail. Is there a way to use the
> '-usesqform' instead? Alternatively, can I do this afterwards and
> combine multiple sessions in a higher-level analysis without feat
> nagging about not having registered properly?
>
> Thanks in advance,
> Cris Lanting
|