Dear Mahmoud,

One thing I would like you to please clarify is this: If someone is interested in comparing motion between two dMRI groups what's the best current way?

I followed all the emails and I'm confused. There is no doubt that the newest eddy does a superior job compared to previous versions of the eddy. But what we discuss here is not to verify the correction ability of new eddy.
Specifically, I want to show the readers that there is X amount of motion between dMRI of two groups before any correction ( so I have to calculate the amount of motion before correction) and then use new eddy and its outputs to show the readers that although there was X amount of motion before correction, after applying the new eddy, we achieved this amount of motion residuals so new eddy did a great job!

Please correct me if I'm wrong. My understanding is that the *.eddy_parameters gives the motion parameters after all correction applied inside the new eddy. If this is correct, then using the *.eddy_parameters is not meaningful to show the initial motion difference between two groups.

No, .eddy_parameters shows what “eddy thinks the movement was before any correction”. After correction eddy “thinks” there is no motion left. This is how all motion-correction/registration works. If eddy (or any other method) thought there were movement left after correction, and it thought it knew what that was, it would correct for it.
 
I'm not sure whether or not *.eddy_restricted_movement_rms gives the motion rms values before or after corrections? Although, the motion rms could be used as a motion index, what if someone wants to have a more detailed look at the motion? let's say wants to know what's the difference in motion in a certain direction?

The .eddy_restricted_movement_rms is eddy’s estimate of what it thinks the movement was before correction. As for “detailed” motion you can use .eddy_parameters, with the caveat I have described earlier that the exact same transform will correspond to different movement parameters for different softwares (that use different methods to do mov_par->transformation_matrix).

My personal opinion is that the movement_rms is the best summary description of how much a subject moved. 

Jesper


I really appreciate if the experts provide a clear and terminal answer to my questions.

Thank you all!
Mahmoud



On Thu, Sep 22, 2016 at 11:49 AM, Jesper Andersson <[log in to unmask]> wrote:
Dear Michael and Thais,


this paper pertains to a method (eddy_correct) that we have recommended against for the last three years. We believe that our current method does a quite good job at estimation movement in the presence of distortions, as can be seen for example in https://www.ncbi.nlm.nih.gov/pubmed/26549300 (which we didn’t write). You can also have a look at figure 7 in http://www.sciencedirect.com/science/article/pii/S1053811915009209 (which we did write).

As has been discussed earlier on the mailbase there is a particular “problem” with estimating translations in the PE-direction, so these should be treated with some caution. Our current recommendation for assessing overall movement is there for the .eddy_restricted_movement_rms. There are some explanation how to use it in the EDDY wiki pages.

Jesper



Did you acquire multiple Bval=0 scans in the course of each DTI? They'd be less distorted by eddy currents, so give you a handle on overall motion during the series.

On 22 September 2016 at 15:18, Thais Minett <[log in to unmask]> wrote:
Dear FSL experts,

I am submitting my DTI-TBSS analyses paper for publication. One of the reviewers asked me: "The authors should report on some metric describing the level of head motion in each group during the DTI scan.". I have no idea how I should address that. I used ‘eddy’ and not ‘eddy_correct’ for my data processing.

Regards,

Thais Minett



--
Michael John Knight, PhD
Elizabeth Blackwell Institute Early Career Fellow
School of Experimental Psychology and School of Clinical Sciences
University of Bristol
Tel 07920 113060