Hi All,
We have a coregistration problem, with a possible solution we wanted to run
by others before implementing. We are acquiring functional images using an
inverse spiral sequence with several groups of older people (patients and
controls, all over 65) on a 4T GE scanner. The problem we have encountered
is that during coregistration of the functionals to the T1 anatomicals, the
functional images are displaced inferiorally and pitched forward (sometimes
as much as 45 degrees) when using SPM2's cost function. Any attempt at
manually re-orienting the images prior to coregistration has not solved the
problem. This problem is carried over during the normalisation step (and
possibly slightly worsened) b/c we determine the normalisation parameters
from the anatomical images and apply these to the functional images. Our
proposed solution is to coregister our anatomical images to the MNI
template, and our (inverse spiral) functional images to the EPI template.
This step places both the anatomical and functional images into the same
space. We then proceed with the normalisation step as described above. Our
first concern with this solution is that it is not standard to coregister to
the separate templates as a means of coregistering to each other (we're
assuming that the T1 MNI & the EPI template are in the same space). Can we
go against the standard in this way? Second, we are wondering if there is
any concern with coregistering inverse spiral images to an EPI template (we
think there could be a concern with normalising inverse spiral to EPI, but
what about coregistering?). What do people think about our problem, and
possible solution?
Just as an addendum while we're asking questions...when we get to the
normalisation phase, we suspect the list's advice will be to create our own
template from a representative sample of controls and patients, as John has
previously advised (given we're working with older subjects, some of whom
have a neurodegenerative disease). The problem we have with this is that we
are conducting a long term study with ultimately around 100 subjects, with
samples of data being analysed along the way. If we analyse a sample of 10
patients and 15 controls, we assume the template should be weighted
accordingly. If we later analyse a sample of 20 patients and 25 controls, do
we have to create a new template with this new weighting, and therefore
pre-process everyone again? Is there a general opinion on how many subjects
are required for a custom-made template, ie when we finally get to 100
subjects, would you want all of these subjects to be used for the template?
thanks,
Melissa & other confused lab members
___________________________
Melissa J. Slavin, PhD
Postdoctoral Researcher
Dept of Radiology and
Brain Imaging and Analysis Center
Box 3918
Duke University Medical Center
Durham, NC 27710-918
Ph: 919-684-7463
Fax: 919-68-7033
|