Dear experts,
We are in the process of analyzing the shape of the subcortical structures in a large, population-based cohort. Initially, we are interested in analyzing sex differences, but later plan on investigating the effects of other covariates. Our volumetric analyses indicate that there is no difference in the volume of the structures. We have some questions regarding the proper way to setup the analysis in FSL. Our research question is whether there are any sex differences in the shape of the structures, despite similar overall volume. As far as we can tell, there are several ways we can setup the analysis (after running run_first_all and concatinating the bvars files):
1. Design matrix with one EV with +1 for men and -1 for women. Run the command "first_utils --vertexAnalysis --usebvars -i structure_name.bvars -d MF.mat -o structure_name --useReconMNI". According to the user guide, the --useReconMNI normalises for brain size.
2. The same as option 1, but also including the --useRigidAlign option. We are unsure how to interpret this option vs. option 1. Should this be included when investigating shape differences between men and women?
3. Design matrix that in addition to sex also includes ICV (and possibly age) as covariates. Should we then run --useReconMNI or --useReconNative (as adding ICV as covariate corrects for head size)?
Following these steps we will run the randomise command with the TFCE option to control for family-wise error rate and are only interested in the t-contrast (+1 for men and -1 for women): "randomise -i structure_name.nii.gz -m structure_name.nii.gz -o MF_structure_name_rand -d MF.mat -t MF.con -D –T"
Any thoughts on which option is best suited at answering our research question?
Best,
Carl Pintzka
|