Dear Jesper,
thank you for your answer. I see that my problem could be that the
activated areas have a large magnitude and spatial extent and therefore
influence the average across all voxel values.
Maybe the scaling should not use the voxel values of all brain voxels
for computation of the scaling factor, but only voxels from CSF, where
we would not expect any useful signal changes. This would imply doing a
segmentation of CSF before scaling.
Regards
Thomas Stephan
Jesper Andersson wrote:
>
> Dear Stephan,
>
> >Dear list,
> >
> >in a block design study with two stimulation tasks and one rest
> >condition we find negative signal changes attributable to one of the
> >tasks in a few cortex areas. In addition the whole CSF (not only parts of it)
> >shows significant negative signal changes.
> >Did anybody find something similar to this,
> >and is there an explanation why the CSF should be less active under
> >the task condition ?
> >
>
> No, I don't think there is an explanation why CSF should be "less active"
> during tsak condition. I think what you see may be an artefact from the
> global normalisation. By normalising the data with the global activity
> (estimated as the average across all intracerebral voxels) then by
> construction we constrain the estimated effects to sum to zero across all
> voxels. This means that if there is an activation somewhere, especially if
> that activation has a large magnitude and/or spatial extent, there will be
> diffuse "deactivations" erroneously introduced elsewhere.
> This problem stems from the average across all voxels not being a perfect
> estimator of some underlying "global" change. Unfourtunately, simply
> skipping the normalisation may also introduce bias. Hopefully there will be
> better ways of performing the global normalisation available in the near
> future.
>
> Good luck Jesper
--
======================================================================
Thomas Stephan Email: [log in to unmask]
Klinikum Grosshadern
Neurologisches Forschungshaus
Marchioninistr. 23 Fon: +49 089 / 7095-4819 *NEU*
81377 Muenchen Fax: +49 089 / 7095-4801
======================================================================
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|