how is TR length impacting ReHo results?
I have performed resting state fmri with TR 2.0 sec and studied reho changes in the cerebellum.
I am now analyzing imaging acquired before this new sequence(same subjects population type): here the voxel size is the same, but we have half of the total number of runs (99 versus 192 of the previous sequence)and the TR is 5.0 . When I process through 3dReHo these dataset I am not able to find any reasonable (kendal coefficient >0.5) reho activation in the cerebellum even with a neighbor size of 93voxels.
What is your opinion?
thanks a lot
I can offer a few musings, but not only mind-wanderings, really:
Off the top of my head, I can’t think of a reason why ReHo would necessarily be influenced by the TR inherently. The time series are just correlated ‘en masse’ within a region using Spearman rank; if your correlation patterns (e.g., due to seedbased correlation) look similar for each acquisition, then the ReHo should look pretty similar, too. So, to investigate this further, I would look at correlation maps and see if anything looks funny.
There might be byproducts of analysis that would affect this. For example, if data is still being low-bandpassed at the same frequency, not the difference in Nyquist frequency associate with each case (TR is units of ‘s’ here):
f_N(TR) = 1 / (2 * TR), —> f_N(2.0) = 0.25 Hz and f_N(5.) = 0.10 Hz.
So, if LFFs are still being bandpassed at ~0.08 or 0.1 Hz, watch out in the second case!
Note that with ReHo, I might have thought that larger neighborhoods might have lower ReHo, because more and more time series are involved that might not be so related; this would likely depend on the region of the brain (e.g., right by a sulcal/gyral boundary vs at the center of a gyrus).