I'm afraid there is no easy solution here.
The problem is that eddy current distortion and susceptibility
induced distortion don't combine in an easy way and neither
affine registration then unwarping or unwarping then affine
registration will completely remove the distortions.
There is a way of combining these in a large model (see
Andersson and Skare, Neuroimage 16(1):177-199, 2002)
but we don't have an implementation of this.
At present we simply try to correct for the eddy current
distortions by affine registration to the image with
no diffusion gradients. It is true that this image still
has susceptibility induced distortion, and that is the
framework that the diffusion is calculated in. You can
unwarp these registered images (using Prelude and Fugue,
if you've acquired a fieldmap) for a first-order correction
which will help a bit. However, the easiest solution is
to work in the eddy-corrected but susceptibility-distorted space.
A better solution, which constrains the unwarping and
registration on the basis of the physics, including knowing
how the fieldmap varies with head orientation and how to avoid
bias in the registration of weighted and un-weighted images, is
part of our ongoing research.
I hope this answers your question.
All the best,
On 5 Dec 2004, at 05:21, Stamatis Sotiropoulos wrote:
> Hi all,
> I have some DTI scans that I need to register. I have heard that DTI
> imagesb have EPI distortion that has to be corrected before the
> registration. Since I am quite new in the field can anybody tell me
> whether this correction is necessary for better registration? If yes,
> can I do that using FSL?
> Thank you in advance,
> Stamatis Sotiropoulos