> we are intending to explore the influence of genetical polymorphims onto
> regional brain volume, esp. a disease-gene-interaction in a large sample of
> patients and controls.
>
> Here we thought to initially use a GLM in a VBM approach with standard
> covariates to remove large influences on brain volumes and to continue with
> the residuals into the genetical analysis.
>
> There are two questions I have,
>
> (1) Is the an easy way to calculate a map of residuals for each subject
> from the output of the estimation (e. g. ANCOVA model or multiple
> regression model)?
You may need to tweek the stats code a little. SPM does produce some residual
images, but they get deleted after they have been used to estimate the
smoothness. In SPM5b, there is some code in spm_spm.m (around line 914) that
says:
%-Delete the residuals images
%=======================================================================
for i = 1:nSres,
spm_unlink([spm_str_manip(VResI(i).fname,'r') '.img']);
spm_unlink([spm_str_manip(VResI(i).fname,'r') '.hdr']);
spm_unlink([spm_str_manip(VResI(i).fname,'r') '.mat']);
end
If this was commented out, then you would get to keep the residual images.
You may also need to increase the value of the global variable
"defaults.stats.maxres"
>
> (2) Rather independently of this, I would like to understand if Jacobian
> modulation is sensitive for only the non-linear warps or also affine
> transformation?
The Jacobian intensity rescaling (modulation) considers spatial scaling from
the affine as well as the nonlinear components.
> Given a generally smaller brain due to brain size, which
> should come near to an average size after normalization, would the
> potentially massive increase of whole brain volume be reflected in the
> modulated maps?
Yes.
|