I have a scan protocol in which a single set of forward/reverse spin echo field maps was obtained towards the start of the scan session, followed by about 45min of different tasks that all use the same acquisition sequence (which are each fed to a separate afni_proc.py pipeline). I have noticed that, due to minor movements in between tasks, etc., the forward direction field map is not always perfectly aligned with the functional scan that is used as the volreg base (in our case, TR=2 of each task). If I am understanding the scripts correctly, I also noticed it appears when using the standard -blip commands in conjunction with afni_proc.py, the script seems to assume that the blip forward/reverse fieldmaps are already registered to the functional/EPI data and does not apply any type of registration prior to applying the blip warp. My question is two-fold:
Would it make sense in this scenario to linearly align the forward field map to the EPI volreg base (and apply the same transformation to the reverse field map)? If so, what would be the best way to accomplish this (e.g., run some separate alignment command(s) prior to running the main afni_proc.py step and then pass in the pre-registered versions as the -blip images?)
Alternatively, is it simply a bad idea to apply distortion correction when the participant may have moved a bit from the exact head position they were in for the field map acquisitions? Might this do more harm than good? We were advised to collect a fresh set of field maps whenever there is any shimming done or whenever the participant has been completely removed from the scanner, but I wasn’t sure of the impact of more minor head adjustments in the space of one scan session.
Thanks for the bump, Ben. I am sorry for missing on the original post.
Distortions do not track with a person’s head. If a subject tilts their head back, the distortion will not simply follow that movement. It may indeed be altered a bit, but the distortion field is not going to follow the volreg parameters, for example.
The distortions that this is trying to correct are mostly along the phase encode (y) direction of the acquisition slices. Those distortions are indeed affected by the subject, and as the subject moves over time, a single estimate of the distortion field may become less accurate. But keep in mind that while the subject may move, the y-axis of the slices remains fixed.
Since the distortion correction is mostly along that one axis, the distortion correction is actually applied BEFORE volreg, so that the y-axis is not altered by the registration step.
Getting to the actual questions now…
No. Rotating the blip images to match the volreg base would assume the distortions follow the anatomy. But while the estimates may not be perfect, distortions are mostly along the y-axis of the acquisition.
The accuracy of the distortion correction will tend to decay (not so simple) as the subject gets farther from the B0 position. So the basic question is, how much does the distortion field change as the subject moves? If it changes a lot, this could be a destructive operation.
However, if the distortion fields are significantly changing, that would apply to the data if you do nothing to correct it, too. I would think the changes would have to be pretty big for distortion correction to be WORSE than not doing anything. Though that also depends on how strongly the corrections are applied.
We are “actively” working on these things now, so hopefully we will develop a better feel for the interplay between motion and distortion. But that will take some time…
rick
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.