EPI distortion is not really linear, so you can't expect a perfect
correction (the amount of scaling in the AP dimension will vary depending
upon the location in the image). If the contrast of the two images is very
similar, you might be able to get a good distortion correction with FNIRT.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of William Copen, MD
Sent: Monday, November 17, 2008 10:26 AM
To: [log in to unmask]
Subject: [FSL] EPI distortion correction using FLIRT (not FUGUE)?
I have an EPI data set that I'd like to register to a FSE T2-weighted data
set. Both are
T2-weighted images, with similar contrast characteristics, and both have the
same
resolution (256x256 in-plane, 16 slices). The patient moved only a little
between the EPI
and FSE acquisitions. The main problem is that the EPI data are slightly
distorted,
basically just a little compressed in the phase-encoding (i.e.
anterior-posterior)
dimension.
I don't have a B0 map, so I can't use FUGUE to correct this distortion.
What kind of
results can I expect from trying to use FLIRT to correct the distortion and
do a little
motion correction at the same time? It seems to me that, as the distortion
is mostly just
a matter of scaling and translation in the AP dimension, I should be able to
get pretty
good (if not perfect) results from an affine transformation. But I've been
disappointed
with what I have tried so far.
Can anyone suggest optimum settings to use for FLIRT in this situation? Is
there a way
that I can constrain FLIRT's fit so that translation and rotation are
allowed in all three
dimensions, but scaling is allowed only in the Y dimension?
Thanks in advance,
William Copen, MD
Massachusetts General Hospital
|