Both ECC and field map correction result in resampling of the data. With
some significant effort (requiring shell scripting skills), you could pull
out the ECC matrix for each volume and convert the field map to a warpfield
and apply them both at the same time to each volume to do resampling in only
a single step. I'm not clear on the benefits of this though.
If you are planning to do any correspondence between subject's structural
and diffusion images, you will need to do field map correction to get proper
alignment, as the distortions of the diffusion image relative to the
structural image are nonlinear. Also, field map correction will improve the
accuracy of tractography slightly, as the gradient directions are applied in
undistorted space. If you are doing something like TBSS to the standard FA
template, distortion correction is not necessarily needed because you are
already nonlinearly registering the distorted FA maps to an undistorted
template.
If your field map correction does not look good, perhaps you can post your
methods here for checking, including what kind of field map you used and
from what scanner.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Forrest Johnson
Sent: Tuesday, June 22, 2010 11:31 AM
To: [log in to unmask]
Subject: [FSL] DTI/Gradient Field Map Correction
Hello All,
With help from this support forum I have been able to gradient field map
(GFM) correct my DTI data (both before and after ECC). However, I do have a
couple of concerns/questions.
Concern 1. Both GFM corrected images appear to have some interpolation
issues that impact the clarity of the images (they are not as crisp as my
non-GFM corrected images, S1_FA.nii.gz files created after DTIFIT). This is
concerning to me. Specifically, I am curious as to how this might
quantitatively impact my images (two interpolations with GFM correction
versus one interpolation without).
Concern 2. Although my frontal regions look relatively better, my temporal
lobes appear to take a big hit (especially when viewing mean_FA images or
any images following tbss_3_postreg).
Q1. Is it really advantageous to GFM correct DTI data?
Q2. Am I sacrificing the quality of other cortical structures in an effort
to correct for field inhomogeneities in frontal regions?
Cheers!
|