Hi Maxime
> 1) Segmenting: I thought realize 1 segmentation (GM an WM in native space with bias corrected), but read it was better to do twice (one segmentation bias corrected YES, then another segmentation bias corrected NO with new images produced). I find it odd, is it really better ?
If the 'standard' segmentation is failing, this two-step method might
help...I've found this, especially for images that have increased
inhomogeneity, for example, due to having multiple coils. If the
standard segmentation is working ok, though, don't bother changing.
> 2)Calcul TIV: I segment GM and WM tissues. I obtain TIV (GM, WM, CSF) with "calculate raw volumes for GM/WM/CSF" button in VBM5.1 tools. I don't understand how VBM can calcul CSF volumes without CSF segmentation images. Do you confirm i use only GM+WM volumes ?
I'm not familiar with the VBM toolbox; however, for the traditional
TIV estimation you would need the CSF partition as well. I would
recommend writing this out during segmentation. Note that your
results will be affected by what you include as a covariate; for
example, if there are changes in GM or WM (e.g., in patients, or with
aging...), TIV will remain relatively stable because CSF will fill in
the gaps. Thus including total GM, total GM+WM, or TIV, as a
covariate, can give you different results (but all are valid, it just
depends on what you want to control for).
> 3)Regression analysis with White Matter: which is the moment i have to treat WM images differently of GM images ? > I do a specific template for WM volumes OR i do 2 regression (one with GM scans [smmwrc1ms*.nii], another with WM scans [smmwrc2ms*.nii°]) OR i do one regression (GM scans then WM scans in the same order) ?
You should have a separate analysis for GM, and a separate analysis
for WM. Don't mix tissue class types within the same analysis.
(Just out of curiosity, I notice you have smmwrc* images....with 2 m
prefixes. I think the 'standard' DARTEL processing would give you
smwrc* images...mw = modulated warped, and then those smoothed is
smw*. I wonder what processing the extra m is from?)
Hope this helps! Good luck.
Jonathan
|