With nonlinear EPI distortions in your image, it will be impossible to get a
perfect registration with either rigid body (6 dof) or affine (12 dof)
approaches. That is because, as you say, some brain regions will be
stretched or compressed, however these changes will not be global (i.e.
linearly correctable). An easy way to find evidence of this is to look at
the genu of the corpus callosum in a distorted FA image registered to a
structural. You will see that the genues do not line up perfectly. After
distortion correction, however, they will line up. The EPI distortions will
affect the tensors also, and you might get somewhat better results with
distortion corrected images in regions that have distortion. Also, with
distortion correction there will be no offsets between where the pathways
really are and where they appear to be when you display them on a structural
image, or offsets between regions of interest defined on a structural image
and where they end up on the diffusion data.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Anastasia A Ford
Sent: Friday, January 16, 2009 12:39 PM
To: [log in to unmask]
Subject: Re: [FSL] T1 to b-zero registration for tractography
Matt,
I have a quick question about non-linear distortions that you mentioned. How
do these distortions manifest themselves? My FA color and lines look correct
and I do not see any irregular hyper intensities in other data. Also, how
would distortions affect the registration? Do you suggest that portions of
the brain volume might be stretched out/warped in the diffusion data and not
on T1? Is this is the case, then this would ultimately affect tractography
results as well.
Sincerely,
Anastasia
|