Dear Laszlo,
> When normalising PET data not to the template, but to the mean-Image,
> that is created
> in the first realignment-step, i.e. doing a transformation of this
> mean-Image on itself, the
> results are really poor.
> The reason for doing this kind of normalisation is to analyse individual
> volumes of the activated/deactivated clusters in specific regions.
> Has anyone an idea, why the results of the normalisation step are not
> very satisfying?
Following on from Alle Meije's comment, I reread your email and became
confused as to why you are doing this. How does this normalization help
analyze individual cluster volumes? Sorry if I am missing something obvious
here.
For what it's worth, you will have a tough time doing the normalization.
First, as Alle Meije points out, you have to smooth the target image (the
mean) in fact to 8mm. Then you will have to disable the template brain
masking, via the defaults, as the normalization uses a mask in the space of
the template to exclude the scalp. This is probably the issue that is causing
the most problems in fact. Lastly, the normalization contains an automatic
bias to find a zoom / shear value that is within the normal range for a series
of normalizations to the MNI template - and these will of course be different
for your mean/mean normalization, where the expected zooms/shears are zero.
Getting this right I think involves a bit of fiddling with the SPM global
vars,
Best,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|