Hi Leandra,
If you mean skull-stripping with something like BET, then I think the
real danger is losing desired tissue voxels in the brain extraction
process itself (which certainly can happen).
The answer depends on which spm version you're using. With spm99/2
where the registration is to a whole-head template, it does seem that
there may be an advantage to registering to a skull stripped AVG152T1
(such a template is available with FSL). This paper sounds very
relevant to your question:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=16442817&dopt=Citation
(or http://tinyurl.com/j7ejm if that breaks)
For SPM5 the priors are effectively skull-stripped already, and I
think the standard ones should be fine; possibly the CSF prior will
include more cortical CSF than the brain extraction leaves, but this
shouldn't adversely affect the GM or WM segmentations. And if you are
really worried about losing CSF near the dura from your CSF
segmentations then you probably don't want to be using brain
extraction at all!
Hope that's of some help,
Ged.
P.S. I think a more interesting question is the opposite (!) -- is
there any danger in using the standard templates (which only model
GM/WM/CSF and "other") with images which include scalp/neck/etc, since
in this case it seems more likely that the segmentation will include
e.g. scalp as GM (something I've observed on occasion with SPM5).
However it seems to me that decent priors instead of "other" would be
very hard to specify, as this non-brain stuff is very variable between
patients (and scanning FOVs, fat-supression sequences, etc.). Anyway,
just a thought.
LEANDRA URIBE wrote:
> Is there any danger in segmenting brain stripped data with the default templates?
> Are there other templates available that would produce better results?
> Do you have any recommendations on how to successfully segment brain stripped data?
>
> Thank you,
>
> Leandra
>
>
> ----------------------------------------------------
> Leandra C. Uribe
> Chemical/ Biomolecular Engineering
> Johns Hopkins University
>
> [log in to unmask]
> ----------------------------------------------------
>
|