Dear Phillip
Thank you for your comments. I tried your suggestion of running VBM
using the standard MNI templates. Indeed, like the paper you refer to
already predicts, there is not much difference compared to using
customized templates.
[I don't understand why you predicted that using the standard priors is
the most sensitive for detecting hippocampal volume differences?]
For both standard and customized templates I see a similar small
hippocampal volume effect. However, these results are still less
convincing (statistically significant) compared to the situation when I
run "cg_vbm_optimized" with defaults.normalise.write.preserve=1 in the
background and analyse the modulated segmentations from the cg
script(the double modulation as you mention, although it is not really
double modulation since the first modulation of the warped T1 gets
largely undone by the bias correction during segmentation).
But I guess the latter approach is not proper VBM...
Kind regards, Dennis
-----Oorspronkelijk bericht-----
Van: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
Namens Philipp Saemann
Verzonden: zondag 24 juli 2005 0:08
Aan: [log in to unmask]
Onderwerp: Re: [SPM] Christian Gaser VBM scripts with "preserve total"
Hello,
there was a very good and 'practical' study on the effect of customized
templates, optimized VBM (i. e. normalisation of gm to GM template
instead of
using wholehead images) and on the effect of modulation: Keller et al,
NeuroImage 2004. It may be worth looking at this systematic comparison.
In brief, modulation seems essential whereas the creation of a
customized
template and priors did not have a large effect. Isn't it rather
recommended
to use own priors and templates in study populations with brains very
different from the MNI template as subjects with (gross) atrophy or
pediatric
populations?
From the point of view of the mentioned paper it might be worth trying a
more
simple approach, i. e. use GM in native space (segmented using standard
priors) for normalisation onto the standard GM template, apply it on
wholebrain, segmentate using the standard priors and modulate the GM
results.
It would be very interesting to see if also in your study this way is
the
most sensitive method to reveal hippocampal volume differences.
Have you compared this 'standard' way of optimized vbm to your approach
which
used two modulation steps?
The reason to produce customized priors is to achieve improved agreement
of
the variable cortices - to my understanding the shape of the GM
partitions is
the essential information which should be translated into the prior
rather
than than the volume information.
Best regards,
Philipp G. Saemann
Max-Planck-Institute of Psychiatry
NMR Study Group
Kraepelinstr 2-10
80804 Munich
|