Hi,
I'm hoping to use the dmri_motion tool to quantify motion for some diffusion data that did not use the eddy_correct tool. We have registration matrices that we've successfully reformatted, but I've been scanning the FreeSurfer forum and see some conflicting information about whether the input to dmri_motion should reflect registration of each time point to a single reference orregistration of each time point to the preceding time point. i.e.
t0-t1, t0-t2, t0-t3, ....
vs.
t0-t1, t1-t2, t2-t3, ...
What input does dmri_motion expect? I think it's the former but the confusion arises from this post:
...
Of our 4 motion measures, the rotation and translation are computed from the output of the eddy correction (which computes relative displacement from one volume to the next), and the intensity drop-out measures are computed from the raw uncorrected images (because the subsequent corrections will introduce some blurring and might lead to underestimation of the drop-out in each slice). So all measures are computed prior to B0 correction.
But, the contents of the “eddy_correct” command appears to register each time point to a single reference time point. We’re using timepoint 1, i.e. b=0.
From the text of Yendiki et al., 2013, I suspect that dmri_motion takes an input that is each timepoint to a single reference timepoint, but outputs a summary that is based on each timepoint to the immediately preceding timepoint?
Can you clarify what input does dmri_motion expect?
Thanks!
Ruth Carper, PhD
Research Assistant Professor, SDSU