3dClustSim

Hello AFNI experts,

I am using Clustsim for the first time and I would like to make sure I am using it correctly. I have also read several posts on the message board and the AFNI –help files but I have a few questions.

My current steps:
(1) Run the 3dFWHMx code on each participant
3dFWHMx -input errts.Proc_9790_April2019+tlrc.HEAD -mask mask_group+tlrc.HEAD -acf

(2)Take the average of the acf values for the participants included in the group analysis

(3) Run the 3dClustsim code
3dClustSim –acf (avg values) –mask mask+tlrc.HEAD –athr 0.05 –pthr 0.001

Follow-up questions:
(1) When I ran the 3dFWHMx code on a few of my participants, I was receiving reports of “out to radius” values ~15mm. In the examples I have seen online these values are usually ~30. Are the values below reasonable?

ID ACF values Radius
6292 0.83439 2.93607 5.85216 7.01209 15.43mm
8225 0.741657 2.8296 10.8613 7.39096 15.99mm
9164 0.800097 2.79238 7.58264 6.89804 15.28mm
9409 0.737612 2.84261 10.3543 7.39743 15.98mm
9444 0.707152 3.22763 9.68089 8.30243 17.97mm
9784 0.75646 3.04957 6.37843 7.38509 15.86mm
9790 0.881869 2.82603 6.81025 6.79469 14.91mm

(2) Do I create the –mask input for the 3dClustsim code? Is this based on all my subjects or can I use one of my participant’s masks from the afni_proc.py?

Thank you very much for your help.
Tamara

Hi Tamara,

(1) The 15-30 mm value depends on the voxel size and the mask.
It does not use a fixed size. I expect that your voxels are
a bit small. How big are they? Did you use a 4-5 mm blur?

Those values look fine.

(2) Yes, you should make a group mask to apply and run 3dClustSim
with. Perhaps make an intersection or 70% overlap mask of all of
the subject full_mask or mask_epi_anat datasets to use for the
group mask.

Consider the commands in this example of a full analysis:

AFNI_demos/AFNI_pamenc/AFNI_02_pamenc/global_process_commands.txt

There is also a related global_process_outline.txt file there.

Does that seem reasonable?

  • rick

Thank you very much for reviewing the values.

Yes, I used a blur of 4 via the afni_proc.py. Based on my protocol, my voxels for my epi volumes are 2x2x3mm and for my anat volumes are 1x1x1.

For the mask, I came across a few threads indicating the use of 3dmask_tool. I created and ran the following code below on a few participants and this seemed to work well. Is there a reason that the group mask is constructed at 70% overlap and not a higher overlap value?

3dmask_tool -input [subjects results folder] full_mask*tlrc.HEAD -prefix mask_group.7 \ -frac 0.7

Thank you very much,
Tamara

Hi Tamara,

The fractional overlap is up to you. We have often used a pure intersection mask (-frac 1.0). Going lower is reasonable, too. It might mostly depend on how close to the edge you are particularly interested in.

  • rick