Hello.
I have been using 3dClipLevel and then 3dcalc to clip and normalize event-related datasets. This is the 3dcalc command:
3dcalc \
-fscale \
-a detrend+orig \
-b mean+orig \
-expr " (a/b * 100) * step(b-1469))" \
-prefix norm
with 1469 being the value determined by 3dClipLevel. I now have some funny datasets where it appears as though I need to switch the clipping procedure, as the normal command manages to clip out most of the real activation and leaves me with mostly noise outside of the brain. The datasets were collected using spiral, and though there is some convincing activation inside the brain, i am getting a large chunk of external noise in the shape of a circle surrounding the entire spiral collection area, a 'halo effect'. Can someone suggest the correct way to mask off this extra space? I have tried switching the sign in the expression to step(b+1469), but that just reduces the data to nonsense. I have also tried using 3dAutomask, but I gather it uses the same value determined by 3dClipLevel, and the result is a mask of only the 'halo effect'.
Thank you.