Hi FSL team,
I have run eddy (5.0.11) several times on the same dataset. Each of those runs differed slightly; the first used "standard" eddy (eddy; so no outlier replacement, slice-to-volume etc); the second time used eddy with outlier replacement (eddy+OLR), and the third time used eddy with outlier replacement and slice-to-volume correction (eddy+OLR+S2V). When I check the eddy_movement_rms output for each run, it is slightly different. For each participant I took the eddy_movement_rms output and took the average of each column to get a single measure of absolute and relative motion (i.e. the same way as done in eddyqc to get single measures of within-volume motion).
On average across participants, standard eddy has the highest motion estimates, eddy+OLR has slightly lower, while eddy+OLR+S2V has lower still. In other words, motion estimates tend to get smaller as more motion correct techniques are applied. Additionally I found that participants with higher estimates of motion under standard eddy tended to have the biggest decreases in motion under eddy+OLR+S2V (the effect is there with eddy+OLR but not as pronounced); participants with low motion estimates with standard eddy also had a similarly low level under eddy+OLR/eddy+OLR+S2V.
I am not all that sure what to make of this, so I have the following questions.
1. Is this expected behaviour and if so, why might it be occurring? I know that eddy can vary from run to run due to a slight random component (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;1ccf038f.1608) but a) this past reply suggests this randomness will have very little impact and b) my finding seems a bit too systematic.
2. Are any of these motion estimates more "accurate" (i.e. are the estimates with eddy+OLR+S2V better than those with just standard eddy, could the standard version of eddy be overestimating within-volume motion relative to the others)?
3. Is there any more information available about about how eddy_movement_rms is calculated? I know the user guide says it is done by "calculating the displacement of each voxel and then averaging the squares of those displacements across all intracerebral voxels". Does this mean that you take the input DWI and the output DWI and compute the displacements based on the comparison of those images? Or is it something else?
Note for what it matters I also repeated with eddy_restricted_movement_rms and get very similar findings.
Thank you in advance!
Stuart
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|