Good afternoon,
I am attempting to create study-specific priors and templates from our
data. I have used several of the scripts found within these archives to
create these files, but the resulting CSF segmentations are very poor. It
appears that a significant amount of tissue remains posterior to the
occiput, skewing any subsequent calculations.
So, what I have done thus far goes like this:
---created study specific priors, via a script found in message#3364
(2003) based on a template derived from our previously normalized data;
subsequent segmentation yeilded the results described above.
---I also repeated this a second time using non-normalized data
(selecting 'not normalized when prompted, and keeping the default values
for the following questions), with results similar to above.
---When I segment using the default priors, the results look very nice.
Only when I use the study-specific priors do I get the poor results.
Any hints...
George P Thomas Jr
University of Kansas Medical Center
gthomas1 *at* kumc *dot* edu
|