Dear SPM Community,
I am a fairly new user of SPM, so I hope that this is not an overly simple
issue.
When I use the SPM defaults for normalization of PET data, I consistently
get a normalized image which is larger than the template image. If I
coregister the subject's MRI, I end up with normalized images that fit
nicely to the template image. Unfortunately, I will not have structural
MRIs for every study and I would like to resolve this problem. I am
outlining in more detail below, the procedures I normally follow and some
attempts I have made to correct this problem.
I am working with PET data collected from Siemens 951 and 921 scanners. I
have been starting with 16bit (int16) ANALYZE format scans, scaling the
scans using MRIcro (v1.27), then setting the origins in SPM using the HDR
edit button. I then realign the images using the default realignment
parameters and create a mean image. I then stereotactically normalize the
mean image to the SPM PET template.
At first I thought that this was a bounding box (BB) issue, since the
default BB for writing the normalized file is a different size than that of
the template. However, changing the default BB to the same size as the
template BB produced only slightly better results. This, coupled with a few
scans which would not normalize, prompted me to experiment with the
normalization parameters. After reading a comment on the SPM e-mail
discussion list suggesting that brain masking during normalization of PET
data may not be appropriate, I tried changing this default for the scans
which would not normalize, but this did not help. I started thinking that
there might be a problem with extra-cerebral voxels in my scans (All of my
scans have negative numbers outside of the head). I decided to make a mask
from the mean image (using the equation i1>0) to use an object mask during
normalization. Upon doing this, the scans which wouldn't normalize before
now would normalize, but were still larger than the template.
The attached PDF file shows a comparison of the PET template to the results
from several different normalization attempts. So far, I have only modified
the brain mask, object mask and bounding box parameters, but none seem to
fit as well as the MRI coregistration. If anyone has any idea as to why
this is happening or has any suggestions for what to try next, I would
greatly appreciate it.
Tim
Tim Ely
Dept.of Psychiatry & Behavioral Sciences
Emory University School of Medicine
1639 Pierce Drive
WMRB Suite 4000
Atlanta, GA 30322
Phone: (404) 727-8264
FAX: (404) 727-3233
email: [log in to unmask]
|