On 02/23/2007 10:16 PM, Joan Fisher wrote:
Dear Caret users:
I have a group of MRIs of children with unilateral brain injury that I
would like to get morphometric measurements on. Specifically, I would
like to know how the gray and white matter in each hemisphere compare
overall (for the whole hemisphere) and how they compare for specific
regions (especially the traditional language areas and the white
matter tracts feeding these areas).
Can Caret give me estimates of the amount of gray and white matter in
each hemisphere (and/or for each region) with patients that have brain
lesions?
Not really. Caret generates a midthickness or layer 4 segmentation
-- midway between the gray/white boudary and the GM/CSF boundary. I
know of no way to generate GM or WM volumes with Caret/SureFit's default
outputs.
FreeSurfer generates both boundaries, and it should be possible to get
volume estimates; I think it even gives you these by default.
Would it be feasible to get this data whether the patient has
periventricular lesions (that don't affect the cortex) as well as
patients where the lesions do affect the cortical surface?
There will be issues with the lesions, but I'm under the impression it
can still be done using FreeSurfer.
With Caret, you can still segment lesioned brains, but you'll probably
need to do manual patching in the area surrounding the lesion.
Finally, if I were to draw the lesioned area (as I have already done
in AFNI, for example), would Caret be able to normalize the brains
around the lesion (according to an atlas or averaged brain)? Or do
you have your own way of handling lesioned brains?
A few people here at wustl.edu have segmented and registered lesioned
brains using Caret/SureFit. There's no standard recipe, and it's
lesion-specific. Some are more troublesome than others. One idea
Donald McLaren (now at wisc.edu) had was to draw a border around the
lesion and use this during registration to contain it. This seems
pretty reasonable to me, but you need a semi-principled way to draw the
corresponding target lesion region on the PALS_B12 target (probably
using the very inflated surface, but with the average fiducial loaded in
a window 2 for reference).
I have had a terrible time finding a method to analyze the structural
MRI information in these brain-lesioned brains, so I would appreciate
your input on this issue.
You're not alone, if that's any consolation. ;-)
Many thanks,
Joan Fisher
University of Chicago
Donna L. Dierker