If you are using distortion corrected DTI data, you should only be using 6
degrees of freedom to register diffusion data of a subject to that subject's
structural scan. I personally have found that affine 12-parameter
registration of distorted DTI data to a subject's structural scan can
produce misalignments in structures that previously were aligned as FLIRT
attempts to correct nonlinear distortions linearly. Also, I have had these
kind of issues of having tractography results displayed outside the brain
when registering the b0, rather than the FA to the T1, and as a result never
use the b0 to register to the T1 any more. The FA is a more similar image
to the T1 than the b0 is, and thus it intuitively makes sense that the
registration will work better. If you have distorted DTI data and don't
want to use FNIRT to register the FA map to the T1, I would try registering
the FA to the T1 with FLIRT and see whether, for your data, 6 or 12
parameters provides better alignment.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Anastasia Ford
Sent: Wednesday, December 31, 2008 12:12 PM
To: [log in to unmask]
Subject: [FSL] T1 to b-zero registration for tractography
Hello and Happy New Year to all!
I have a question regarding registration between the b-zero diffusion volume
and the T1
anatomical scan. Generally, we draw our ROIs for tractography on the T1 scan
and then use
the transform provided by flirt to register these masks to the diffusion
space. However,
when tracts are displayed in the T1 space, they often tend to transverse the
ventricles and
project outside of the brain volume. We have tried changing the degrees of
freedom in flirt
to see whether this would improve the registration, but it seems that the
registered volume
appears to be larger than the diffusion volume, that is, the T1 brain
registered to b-zero is
stretched out beyond the borders of the diffusion volume. Are there any
additional ways to
improve the linear registration, besides the change in degrees of freedom?
What is
considered to be the ideal setting for the degrees of freedom for the best
registration
results?
Sincerely,
Anastasia
|