Dear all,
I'm puzzled by some peculiar results we found using FIRST.
We used FIRST (as implemented in run_first_all) to a set of healthy controls and patients with multiple sclerosis (T1 weighted images 1mm isotropic).
We performed volumetric analysis and shape analysis (with 6DOF reconstruction) of the subcortical nuclei.
Volumetric data were submitted to an ANOVA (group, hemisphere, nuclei,with the TIV as nuisance variable) while shape analysis was performed separately for each nucleus with a two sample t test with the TIV as nuisance variable (with randomise).
We found that volume was significant different between the two groups for caudate nucleus, putamen and others structures, and the shape analysis revealed consistent clusters of local atrophy in these structures.
At the same time, volume in the amygdala and in the pallidum was not different between the groups, as revealed by the ANOVA, but the shape analysis revealed huge (i.e. a unique cluster spanning the entire perimeter) cluster of local atrophy in these very same nuclei.
I can't really understand how is possible that a "global local atrophy", as revealed by the shape analysis, is not mirrored by a global atrophy as revealed by the volumetric analysis.
Are the scales of volumetric analysis (mm3 or voxel) and shape analysis (??) completely different ?
Is it possible that consistent changes in some local value measured by the shape analysis is not reflected in more global analysis ?
I hope I made myself clear, any opinion, tip and comment is more than welcome.
Alain
|