Motion Sensoring

AFNI Message Board
1

Hi all,

I'm preprocessing resting-state fMRI data from a clinical Parkinson's disease (PD) cohort using afni_proc.py and would appreciate guidance on an appropriate motion censoring threshold.

My current pipeline uses:
-regress_censor_motion 0.3
-regress_censor_outliers 0.05

A colleague raised the concern that 0.3 mm may be too strict for a PD population, where tremor and motor symptoms naturally increase head motion compared to healthy controls.

My cohort details:

  • Clinical PD patients, pre-operative DBS cohort
  • Two sites: site 1 (~152 subjects, GE 1.5T and 3T, Siemens 3T) and Site 2 (~100 subjects, uniform 400 TRs, TR=0.75s)
  • No fieldmap/reverse phase encoding available
  • Primary ROIs: deep brain structures (striatum, thalamus, pallidum)
  • AFNI 26.1.02, using enorm for motion censoring

My questions:

  1. Is 0.3 mm enorm too strict for a clinical PD population? What threshold would you recommend?
  2. Are there published resting-state fMRI studies in PD or other clinical/elderly populations using AFNI that you would point to for guidance?

Thank you!

2

Hello,

Yes, that censor level would need to be appropriate for the population. A reasonable number can also depend on the scanner setup (comfort, stimulus types, etc).

It is common for people to use 0.5 mm enorm for higher movers, and in some cases 0.7, 0.8 or even 1.0. The 0.3 mm we use in many examples is for healthy adults in a task design, 0.2 mm for rest.

Historically we might suggest getting rid of as many bad time points as is reasonable, while maintaining enough data for the study. Many people are drifting toward doing less censoring (higher threshold).

To evaluate the censor counts for various censor levels, consider looping over something like:

1d_tool.py -infile motion.1D -set_nruns 9      \
           -show_censor_count                  \
           -censor_motion 0.25 sub-219

for various subject motion.1D files and various censor levels.

Does that seem reasonable?

-rick

1 Like