Extremely Large Euclidean Norms

Hi all. Wanted to see if anyone else had run into this before.

I used AFNI’s 3dTo1D to calculate the euclidean norm, expecting it to be RELATIVELY similar to FD. However, the values I’ve gotten for the output are EXTREMELY high.

Range is from

This is SMS data, so I’m aware it should be notch filtered prior to any nuisance regression, but I am not doing so because what I’m examining is the relationship of the motion/respiration to independent components.

I’d like to know if I’m reading this right, because it looks like I have pretty large motion differences between my two conditions.

Has anyone encountered this before? Or is this a massive scaling SNAFU on my part?

These are the commands I used:

3dTto1D -input sub-${subjectID}/Session1/func/closed/sub-${subjectID}_Session1_task-restclosed_bold.nii.gz -method enorm -mask sub-${subjectID}/Session1/func/closed/Masked_Skullstripped_SBref_sub-${subjectID}_Session1_task-restclosed_bold.nii.gz -prefix $motiondir/sub-${subjectID}_ENORM.1D

1dplot -volreg -sepscl -jpg $motiondir/sub-${subjectID}_ENORM.jpg $motiondir/sub-${subjectID}_ENORM.1D

Thanks in advance.


This seems to happen regardless of what method I use. Ended up with DVARs values in the hundreds.

fsl_motion_outliers doesn’t seem to be having this issue so it’s likely the way I’m passing the data to 3dTo1D.


You are currently computing the Euclidean norm of the EPI data, but probably want to compute it from the 6 motion parameters. Do you have a text file of the 6 motion parameters? If so, 1d_tool.py can be used to compute the enorm from those. See example 9 from “1d_tool.py -help”.

3dTto1D can compute the enorm of the motion parameters, too (see example E3), but 1d_tool.py is a more general tool for data such as motion parameters.

  • rick