Image registration in SPM involves searching for a set of parameters
that optimise some registration criterion. The reason for manual
re-positioning is to improve the chance of the initial affine
registration finding a globally optimal solution - rather than a
solution that is locally optimal. Providing the starting estimates for
the parameters are within a few cm, and not so many degrees, then the
affine part should hit on the globally optimal solution. If the exact
same match is reached for this affine registration, then the final
registration result should be identical - whether or not the image has
been manually moved (providing there was no reslicing involved).
Your strategy of using the so-called "optimized VBM" strategy for
spatial normalization is a good one - although if you use the
Segmentation in SPM5 to estimate your spatial normalization parameters,
then this should do pretty much the same (although it also
simultaneously matches WM). See e.g.
Spatial normalization of lesioned brains: Performance evaluation and
impact on fMRI analyses
NeuroImage, In Press, Accepted Manuscript, Available online 24 May 2007,
Jenny Crinion, John Ashburner, Alex Leff, Matthew Brett, Cathy Price and
Karl Friston
If you are using SPM, and the anatomical and functional data are well
registered, then you should be able to apply warps estimated from the
anatomical images, to your fMRI. If your functional and anatomical data
are not well aligned, then you have no chance of doing this.
Best regards,
-John
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
On Behalf Of Jeff Browndyke, Ph.D.
Sent: Monday, June 04, 2007 4:36 AM
To: [log in to unmask]
Subject: [SPM] Mod of warp transform?
This issue has come up in a couple of labs here, and I thought it
warranted
feedback from the SPM group. We've run into problems using the SPM
co-registration process with patients having moderate or greater
atrophy...most often the result is a large difference in the pitch
between
the functionals and anatomicals. So, this led us to wonder if one
manually
reorients both the patient functional (3.75mm isotopic) and anatomical
(1mm
isotopic) to the same origin in SPM, such that pulling up the resulting
images in "Check Reg" accords good overlap, does this pose a problem for
subsequent transformation? I suspect so, but just wanted to
double-check.
Our current scheme is to segment the anatomical, then warp the grey
segment
to a grey atlas space. The resulting warp transform parameters are then
applied to the functionals. Without prior co-registration, I would
think
that the transform parameters based upon 1mm isotopic could not be
directly
applied to 3.75mm functionals without serious error.
Since SPM co-registration is causing us problems with some subjects and
our
functionals and anatomicals are reoriented and resliced to a common
origin,
is there a way to mathematically change the 1mm-derived warp
transformation
parameters to fit the 3.75mm functionals? Essentially, we would like to
skip co-registration while still correcting for voxel dimension
differences
and using this step-wise warping approach.
Regards to all,
Jeff Browndyke
|