This was asked via e-mail and might be of general interest.
> After 3dDWItoDT, I can get 6 components tensor data, 3 lambda data, 3
> eigenvector data, FA, and MD.
> Is there anyway to generate relative anisotropy (RA) map (which I can
> see from DTIstudio)?
The relative anisotropy, RA, is a scaled version of the
standard deviation of the eigenvalues.
RA = sqrt( ((L1-Lm)^2 + (L2-Lm)^2 + (L3-Lm)^2)/3 ) / Lm
where L1, L2, L3 are the three eigenvalues, and Lm is the mean
eigenvalue.
% 3dcalc -a ks_linear2_L1.nii -b ks_linear2_L2.nii -c ks_linear2_L3.nii \
-expr 'mean(a,b,c)' -prefix ks_linear2_Lmean.nii
% 3dcalc -a ks_linear2_L1.nii -b ks_linear2_L2.nii \
-c ks_linear2_L3.nii -d ks_linear2_Lmean.nii \
-expr 'sqrt(((a-d)^2 (b-d)^2 + (c-d)^2)/3) / d' \
-prefix ks_linear2_RA.nii
% 3drefit -sublabel 0 RA ks_linear2_RA.nii
Not surprisingly, the results look very similar to the FA.
Qualitatively, they are about the same although the values at each voxel
differ. There are other anisotropy indices that could be calculated like
the lattice index and the coherence index, but they are somewhere low
down in the priority list. Simple ones like this can be calculated using
3dcalc commands fairly easily.
