On 2/14/19 1:04 PM, John Absher wrote:
External Email - Use Caution
Hi, I realize this is fairly straightforward for most of you, but I am relatively new to freesurfer and want to make sure I’m proceeding correctly with group analysis.
We have ~35 TBI subjects and a similar number of age- and sex-matched controls, and we’re trying to see if we can identify differences between groups in cortical thickness, curvature and volume. We’re using parasagittal T1 post-contrast 3D volumes obtained as part of routine clinical care, and hope our data show that it is feasible to do retrospective analysis like this, when axial 3D T1 MRI volumes are unavailable.
So we have:
Nclasses=4 (TBI+, TBI-, M, and F)
Nvariables=1 (age)
NregressorsDODS = 8
Each of the contrasts is repeated for lh (left hemisphere) and rh (right hemisphere), and the 3 values for MeasurementName. The “Title” in the fsgd file is changed to specify a separate output directory for each contrast. We figure there should be 24 lh directories, and 24 rh directories when we’re all done, with 8 contrasts (below) for thickness, volume and curvature (8*3), for each hemisphere (i.e., 48 total).
The contrast names and matrix specifications are:
Male.TBI-vs-Controls.intercept.mtx 1 -1 0 0 0 0 0 0
Male.TBI-vs-Controls.slope.mtx 0 0 0 0 1 -1 0 0
Male-female.intercept.mtx 0.5 0.5 -0.5 -0.5 0 0 0 0
Male-female.slope.mtx 0 0 0 0 0.5 0.5 -0.5 -0.5
TBI-Controls.intercept.mtx 0.5 -0.5 0.5 -0.5 0 0 0 0
TBI-Controls.slope.mtx 0 0 0 0 0.5 -0.5 0.5 -0.5
Gender-x-TBI.intercept.mtx 0.5 -0.5 -0.5 0.5 0 0 0 0
Gender-x-TBI.slope.mtx 0 0 0 0 0.5 -0.5 -0.5 0.5
Should we do other contrasts?
It is totally up to you, but I don't think there are any other interesting contrasts.
Female.TBI-vs-Controls.intercept.mtx
Female.TBI-vs-Controls.slope.mtx
Age-x-TBI.intercept.mtx
Age-x-TBI.slope.mtx
Finally, does a permutation analysis need to be done once, or once for each contrast?
If you run mri_gmfit-sim with the permutation option, it will do all the contrasts at once.
Here are the first few lines of the FSGD file:
GroupDescriptorFile 1
Title lh.gender_age.glmdir
MeasurementName thickness
Class TBIMale plus blue
Class ControlMale square red
Class TBIFemale circle pink
Class ControlFemale black triangle
Variables Age
Input TBI1_002 TBIMale 45.000
Input TBI1_003 TBIFemale 46.000
Does this approach look correct? Apologies for the long post.
Thanks,
John
*John R. Absher, MD, FAAN*
Department of Internal Medicine, Division of Neurology
Prisma Health–Upstate
200 Patewood Drive, Suite 350B
Greenville, SC 29615
864-454-4500 (office)
864-350-6655 (mobile)
864-454-4505 (fax)
Inspire health. Serve with compassion. Be the difference.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer