Distortion correction for combined brain + cervical cord fMRI

MRI Acquisition
1

Hello everyone,

My background is mostly brain fMRI, where acquiring a reverse phase-encode pair or a field map for distortion correction is fairly standard. I’m now working with 3T fMRI that covers the brain and cervical cord in one acquisition. We currently don’t acquire anything for distortion correction. I’ve looked through the literature but couldn’t find much clear guidance for the cord. I also saw that this was discussed here for diffusion data. Have there been any updates since then, particularly for fMRI? I’d love to hear how others approach it.

A few things I’m curious about:

  1. Reverse phase-encode (topup) or field map: which has worked better for you in the cord, given its small size and the CSF, bone and air interfaces in the neck?
  2. Brain only: if we can’t get a usable field estimate for the spinal cord, does it make sense to correct only the brain portion? Or does correcting one region and not the other cause more problems than it solves?
  3. Sequences: are there sequences or shared protocols available that are optimized for distortion mapping in the cord?
  4. Is it worth it at typical cord fMRI resolutions, or have you found it makes little difference?

Any experiences, papers or example pipelines would be much appreciated. Thanks!

2

Hi @pkotlarz,

Here’s a relevant discussion on that topic: How to correct for distortions in spinal cord diffusion MRI data?

It’s pretty much the same context as both diffusion and functional MRI are (mostly) based on EPI readout and hence suffer from similar geometric distortion.

One slight difference thought, is that since fMRI protocols use T2*w, they also suffer from signal dropout, which makes the registration-based method trickier to implement. So far, the community has mostly been relying on segmentation-driven non-linear registration (between the average EPI time serie and the anatomical scan), which obviously requires a fairly accurate spinal cord segmentation (which is anyway required for registration to an atlas).

Cheers,
Julien