That looks right, though I'm not sure what is in the design matrix. Is it a column of +1 and -1?
On 11/9/17 12:29 PM, Mehta, Chintan wrote:
Dear Doug,
I think that is essentially what I figured. Based on remarks in other posts, my steps for evaluating laterality index entail applying:
(1). Apply mri_preproc for each subject $SUB and $meas (area or thickness) individually using -xhemi flag but without any paired differences via:
mris_preproc --target fsaverage_sym --hemi lh --xhemi --srcsurfreg
fsaverage_sym.sphere.reg --meas $meas --out ${SUB}_out.mgh --s $SUB Result is a surface file with two volumes (lh and xhemi-lh) called ${SUB}_out.mgh for each subject.
(2). Smooth the two volumes in ${SUB}_out.mgh at a given $fwhm (5, 10, 15...):
mris_fwhm --s fsaverage_sym --hemi lh --cortex --smooth-only --fwhm
$fwhm --i ${SUB}_out.mgh --o ${SUB}_out-sm${fwhm}.mgh to create a surface file with two volumes whose values are smoothed at fwhm = $fwhm.
(3). Compute laterality index over the surface for each subject via paired-diff-norm:
mri_concat ${SUB}_out-sm${fwhm}.mgh --paired-diff-norm --o
${SUB}_latIndex-sm${fwhm}.mgh where ${SUB}_latIndex-sm${fwhm}.mgh is the surface file with laterality index for subject $SUB (after smoothing measures over left and right hemispheres).
(4). Stack laterality indices for each subject in sample G (of n subjects) into a single mgh,
mri_concat --i SUB1_latIndex-sm${fwhm}.mgh --i SUB2_latIndex-sm${fwhm}.mgh .......
--i SUBn_latIndex-sm${fwhm}.mgh --out --i sampleG_latIndex-sm${fwhm}.mgh.
(5) Run glm with designed matrix sampleG_design.txt containing n rows of covariates for SUB1, SUB2, ..., SUBn (in order) via:
mri_glmfit --y sampleG_latIndex-sm${fwhm}.mgh --X $sampleG_design.txt --no-rescale-x
--glmdir glm_sampleG --C con1.mat --surf fsaverage_sym lh
The whole purpose of this exercise + question was that I need to fit the GLM over subsets of the sample for sensitivity analysis to determine whether significance for the contrast in con1.mat (say slope of an IV, such as age) depends on patterns of based on demographics groups (such as gender). I suppose alternatively, I could just as well fit another GLMs with a design matrix modeling interactions between age and gender to make my evaluation, over the same set of stacked surfaces for the full sample.
Thank you.
Best, Chintan
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer