How should the censor_motion parameter be set in 1d_tool.py, and what is a reasonable value for it

AFNI Message Board
1

AFNI version info (afni -ver): AFNI_24.0.04 'Caracalla'

Dear all,

I am using AFNI to censor my frmi data.
And I do not know how should I set a reasonale value for censor_motion in 1d_tool.py.
My fmri data comes from task-based mice experiments. (TR= 1s, total 1000TR scanned)
When I set the censor_motion below 1 mm, all the total numbers will be censored.
When I set the censor_motion equals to 1.5 mm, 156 TRs are censored.
When I set the censor_motion equals to 2 mm, 40 TRs are censored.

I am really confused what will be a suitable parameter for censor_motion.

Thanks for your reading!

my code is below

code text  # or delete if not needed
1d_tool.py -infile de_mot.txt -show_censor_count -censor_prev_TR -censor_motion 1.0 censor.1D
2

Hi-

Firstly, I don't think there is a universally established censor limit.

However, values that are often used in studies typically start with something like 0.2 mm (for resting state, human adult subjects that might not be very likely to move), to 0.3 mm (task based FMRI), to perhaps a bit higher up to 0.5-1 mm (more motion prone subjects, or with a task where movement might occur, particularly in task-based FMRI). There are tradeoffs of sensitivity and specificity here: the higher the threshold, the more data and time points you keep, but also the more effects of motion are in your data. This is a large topic of research in the field still to understand.

With mice, whose brains are understandably pretty small, 1 mm of motion seems quite large already. I'm guessing that the voxel sizes are quite a bit smaller than 1mm, yes?

Have you tried processing your data with afni_proc.py? We have applied it successfully to make animal datasets, including macaque, marmoset, rats and dogs.

And have you looked at the raw data, to make sure that it really looks like real motion is present? Sometimes, animal scanners have brightness fluctuation patterns over time, and that can make it appear like there is motion where isn't any (or at least make motion appear to be much larger than it really is).

--pt

3

Hello,

Thank you very much for your detailed explanation.
I got your point. Meanwhile, it reminds me that when i doing the preprocessing, I enlarged the mice NIFTI files for 10 times, since mice's voxel sizes are quite small.
afni_proc.py seems useful, I will try to learn it later.

Thanks again!

Have a good day ahead!

XINGYU

4

If you use @animal_warper for alignment to a template, you don't have to fake a voxel size 10x larger. We provide a version of the Allen mouse brain template and atlas here:

AFNI_AllenMouseCCF3.tgz

and the older version of that atlas here:

5

Thank you so much for your advice!
I will try to use Allen mouse brain template and atlas.

--XINGYU