John:
OK I did some additional tests with the various modules and their behavior
is not quite what I had expected. But I think (hope) I now understand the issues.
Let's say I start with the SPM96 template. It's native transformation matrix
is:
2 0 0 -92
0 2 0 -128
0 0 2 -74
0 0 0 1
Now I add a 90 deg roll to this matrix and get.
0.0000 0 2.0000 -74.0000
0 2.0000 0 -128.0000
-2.0000 0 0.0000 92.0000
0 0 0 1.0000
In order to make this work in spm_coregister the image defining space has to
be an image with the original transformation matrix. Then my image
is rotated 90 deg as expected. I guess that's what you mean by image
defining space, but for some reason that never clicked before. I just
thought it wanted the basic header info.
Now for spm_realign I told it to reslice images. I picked my one image and it
does not seem to incorporate the transformation. (My original post about
this was in error here. It doesn't use the transformation matrix at all, not
just for rotations, if you only feed in one image). I guess what I thought it
would do is just assume a transformation matrix based on header info. But it
seems that the routines don't know how to make the transformation without
some baseline image, even when given the *.mat files.
This brings me to a new scenario.
I have a set of images which I realign to an image called Q. Q might be the
last image of a set of functional images that was taken right before the T1.
I start out by only performing a realignment to Q and not reslicing. Then I
ask spm to normalize and reslice the images. BUT this time I don't include Q
as the first image in the set of images to normalize. This is going to cause
a problem isn't it, because the transformation matrices were all made
relative to Q, and now without Q in the set of images to reslice the
appropriate reference isn't there?
What I thought spm was doing was assuming some unrotated form of the image to
reslice as a baseline, but it seems to actually need the
unrotated/untranslated, etc. matrix as part of the initial image (or the
matrix of whatever image was first).
Several spm answers allude to the fact that one can realign to a particular
image but then not use that image when reslicing and I'm not sure this is
correct.
http://www.mailbase.ac.uk/lists/spm/1997-08/0020.html
as an example.
Please let me know if the scenario involving image Q is correct. What this
would mean then is that in reslicing one would always have to include the
image realigned-to as the first image in the series.
thanks,
darren
--------------------------------------------------------------------------
Darren R. Gitelman, MD Email: [log in to unmask]
Cognitive Neurology and the Voice: (312) 908-9023
Alzheimer's Disease Center Fax: (312) 908-8789
WWW: http://www.brain.nwu.edu
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
--------------------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|