Will you may get a quick telegraphic reply on this, these are each really
separate and quite complex questions and you will likely get many answers
For subcortical volumes, while dividing by TICV has become a common
solution and we have done it, it is not the optimal method. Introduces
fixed error for all structures in a subject. See footnote in Buckner
paper on this. If you are doing between group analysis, analysis of
covariance is preferable. Proc glm or mixed in SAS, easy in other
packages This can handle as covariates ICV, and depending on your
population, you can add handedness, gender, IQ or the kitchen sink.
Another plus for covariance is that you get an interpretable group mean, a
real number that can be compared to the literature instead of some weird
fraction.
For surface, we have used the (third root of TICV), squared, based on
suggestions by Bruce Fishl, I don't see why covariance with this as the
term would not be slightly preferable to diving for the same reasons as
above.
For thickness, two schools of thought. Some do no correction. Others use
mean thickness over the whole brain of the individual or a subregion. It
has been convincingly argued that what you use and whether you use it it
depends again on the question you are actually asking. There are some
very good and smart threads on this if you search. Mike Harms I seem to
recall
Carl Schwartz
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
hi miro,
as per me, i would suggest to divide ICV with TotalGrayVol and other
aseg.stats files' global measures to remove the invariability across these
measures. divide ICV to
totalgrayvol,corticalwhitemattervolume,supratenorialvol,subcorticalgrayvol...this
will remove any in-variability within those measures across ICV......
hope that helps
kk
----- Original Message -----
From: "Miro Drahos" <miroslav.drahos(a)ucsf.edu>
To: freesurfer(a)nmr.mgh.harvard.edu
Sent: Thursday, June 21, 2012 2:21:13 PM
Subject: [Freesurfer] correcting for normal inter-cranial variability
Hi all,
When analyzing output measures from FreeSurfer, what is the proper way to
correct for normal anatomical variability before comparing between
subjects? Which variable should be used to calculate normalized measures?
volumetric measures -- total intra-cranial volume? total brain volume?
surface measures -- total surface area?
thickness measures -- mean thickness?
Thank you!
Miro Drahos
Hi all,
When analyzing output measures from FreeSurfer, what is the proper way to
correct for normal anatomical variability before comparing between
subjects? Which variable should be used to calculate normalized measures?
volumetric measures -- total intra-cranial volume? total brain volume?
surface measures -- total surface area?
thickness measures -- mean thickness?
Thank you!
Miro Drahos