sMRI global volume correction
ab698
Posted on 08/12/10 14:31:12
Number of posts: 18
Do you guys think it is necessary to include the global tot intracranial or tot gM volume as a nuisance covariate or correct the input images by doing a proportional scaling correction.
thanks
anna
Untitled Post
I'm Online
nlobaugh
Posted on 08/12/10 14:58:38
Number of posts: 229
tough call.. it most likely depends on the exact processing steps used to derive the VBM data.
If you have accounted for global differences in brain size during the preprocessing, then using it as a covariate will probably not have a huge effect on your data.
there is a short discussion of this issue in a recent "Pitfalls of VBM" paper by Henley et al:
AJNR Am J Neuroradiol 31:711–19
April 2010
DOI 10.3174/ajnr.A1939
nancy
Untitled Post
ab698
Posted on 08/12/10 15:06:13
Number of posts: 18
Yes exactly that's why I thought I'd try and get a community consensus opinion. I've always felt that when doing Controls vs Patients the global differences can often be important. But anyway we used the FSLVBM pipeline which included the jacobian modulation.
Untitled Post
I'm Online
nlobaugh
Posted on 08/12/10 15:37:29
Number of posts: 229
my gut feeling is that you don't need to do it.. especially since you've already modified the probability of a voxel being GM with the jacobian
As long as the jacobian is derived after the linear transform to template (which accounts for head-size differences), then you only have to deal with the interpretation of what a "modulated" voxel value actually means.. and it can mean many things, since this value is derived by multiplication
you may also want to consider evaluating the jacobians directly - the interpretation will be much more straightforward
nancy