Before moving on to using 3dReHo, we need to first ask some important rhetorical questions (IRQs):
This is demonstrated in the following video, which was recorded shortly before I was assaulted by my girlfriend with dental floss. Unwaxed, of course. (She knows what I like.)
- Did you know that with the advent of free neuroimaging analysis packages and free online databanks, even some hirsute weirdo like you, who has the emotional maturity of a bag full of puppies, can still do FMRI analysis just like normal people?
- Did you know that it costs ten times as much to cool your home with air conditioning than it does to warm it up?
- Did you know that Hitler was a vegetarian?
- Did you know that over thirty percent of marriages end by either knife fights or floss strangulation?
Once we have thought about these things long enough, we can then move on to running 3dReHo, which is much less intimidating than it sounds. All the command requires is a typical resting state dataset which has already been preprocessed and purged of motion confounds, as well as specifying the neighborhood for the correlation analysis; in other words, how many nearby voxels you want to include in the correlation analysis. For each voxel 3dReHo then calculates Kendall's Correlation Coefficient (KCC), a measure of how well the timecourse of the current voxel correlates with its neighbors.
3dReHo can be run with only a prefix argument and the input dataset (for the following I am using the preprocessed data of subject 0050772 from the KKI data analyzed previously):
3dReHo -prefix ReHoTest -inset errts.0050772+tlrc
The default is to run the correlation for a neighborhood of 27 voxels; that is, for each voxel touching the current voxel with either a face, edge, or corner. This can be changed using the -nneigh option to either 7 or 19. In addition, you can specify a radius of voxels to use as a neighborhood; however, note that this radius is in voxels, not millimeters.
E.g.,
3dReHo -prefix ReHo_9 -inset errts.0050772+tlrc -nneigh 9
Or:
3dReHo -prefix ReHo_rad3 -inset errts.0050772+tlrc -neigh_RAD 3
Lastly, if your current dataset has not already been masked, you can supply a mask using the -mask option, which can either be a whole-brain mask or a grey-matter mask, depending on which you are interested in. For the KKI dataset, which is not masked during the typical uber_subject.py pipeline, I would use the mask_group+tlrc that is generated as part of the processing stream:
3dReHo -prefix ReHo_masked -inset errts.0050772+tlrc -mask mask_group+tlrc
This is demonstrated in the following video, which was recorded shortly before I was assaulted by my girlfriend with dental floss. Unwaxed, of course. (She knows what I like.)