Hi Darren,
I don't have very strong thoughts on the issues. However, as SPM allows
a proportional scaling normalisation, I figured that the extra
flexibility to chose whatever globals you like, could be a good thing.
Although many people don't like the MANCOVA approach to correcting for
head size, brain size or whatever, the option is also still available.
Factoring out the volume change from the affine registration will also
be influenced by how the registration was done. Registration may be
based on the whole head, whereas sometimes it may be based only on the
brain. These will give slightly different affine transforms, so
variability in skull thickness etc may influence the results. I'm not
entirely certain what data and objective function is used by the FLIRT
registration of FSL's processing stream.
In some cases though, people may just want to see what is bigger in one
group compared with another. The ability to skip the corrections also
allows this option.
I have not looked into which approach is "best". In part, it depend on
what you want to find out - although there is a component to the
decision that concerns figuring out the best model of the data.
All the best,
-John
On Wed, 2010-05-05 at 09:00 -0500, Darren Gitelman wrote:
> John
>
> These notes are excellent. Thank you.
>
> One question. I notice that SPM modulates by both the affine and
> non-linear components of the Jacobian, so that brain size differences
> are corrected for by including a total intracranial volume covariate.
> Another approach taken by FSL and the VBM toolbox (Gaser) has been to
> modulate by only the non-linear component so that the affine (scaling)
> differences are accounted for and no covariate is needed. Do you have
> any thoughts on why one would choose one approach over the other?
>
> Darren Gitelman, MD
>
>
> On Wed, May 5, 2010 at 8:25 AM, John Ashburner
> <[log in to unmask]> wrote:
> Some of you who were in the VBM groups of the Edinburgh SPM
> course last
> week may want a copy of the notes. Here they are...
>
> http://www.fil.ion.ucl.ac.uk/~john/misc/VBMclass10.pdf
>
> All the best,
> -John
>
> --
> John Ashburner <[log in to unmask]>
>
--
John Ashburner <[log in to unmask]>
|