Hi Daniel,
I might be wrong about this, but I think you're talking about
different things in FSL and SPM...
> Next, they determine the paramaters of the non-linear transformation
Unless you are using Daniel Rueckert's IRTK stuff under the
$FSLDIR/bin/DR directory, then FSL/FLIRT is only doing linear (or
affine if you prefer) registration.
> from the low-quality T1 to the high-quality T1 (eg MPRAGE)
> and from the MPRAGE to the standard-space template. They then multiply
> those two transformation matrices
Indeed, this homogeneous matrix representation is only possible for
linear (or being pedantic, projective) transformations.
> do each transformation and apply the parameters separately, but warping the
> functional data twice will create more opportunities for the data to be
> mishandled; I would prefer a single warp with the combined parameters.
If you perform rigid or affine coregistration or motion correction in
SPM I believe (someone please correct me otherwise!) that the
transformations are stored either in .mat files or in the NIfTI
headers (under SPM5). So the images actually are not
resampled/transformed multiple times; the matrices are combined and
updated, as you suggest.
Non-linear warping in SPM (using a DCT basis) is different to what FSL
does, and indeed, usually does result in new resampled warped images
being produced. However, you can use "normalise->write" to apply a
warp to another coregistered image. So if you do e.g. functional to T1
with rigid coregistration, then warp T1 to template (e.g. with SPM5
unified segmentation), you can transform functional to template using
the same warp, without the functional being first resampled from the
rigid coreg. I hope that makes sense!
If you actually want functional warped (DCT) to T1 [and I don't know
enough about fMRI to know if this is a good idea -- but reiterate that
I don't *think* FSL does this] then this should be possible using the
method Marko suggests. In SPM5 if you click
"Tasks->Util->Deformations" you can create a "composition" of two or
more warps.
The T1->template using unified segmentation would give you a
seg_sn.mat file that you would "import" into the composition. For the
functional->T1 warping I think you'd have to use the HDW toolbox
(Tasks->Tools->High-Dimensional Warping). Unless I am missing some
fMRI motion-correction/unwarping stuff that I know nothing about...
Best,
Ged.
|