Dear Mark,
If the brains are from the same individual then you shouldn't
use more than 6 (or maybe 7) dof for the registration. However,
getting partial-brain EPI matched to full-brain anatomical scans
is quite difficult. We recommend that when you do an experiment
like this you spend another minute and acquire a full-brain EPI
scan (including the same slices as in the partial-brain one, but
also extending both ways to cover the whole brain). This way
the full-brain EPI has the same distortions, resolution and
contrast as the partial-brain one and is much easier to register
with the full-brain version. This kind of registration is easy to
set-up within the FEAT registration GUI, or is relatively easy to
run via the flirt command line too.
If you do not have a full-brain EPI then the problem is much harder
but still achievable. To start with you will probably need to get an
initial alignment (similar to what BrainVoyager does by the sound of
it).
This can be done using the Nudge GUI. Then, once you've found
your initial alignment you will need to run FLIRT with the search set
"Already virtually aligned (no search)" and also use a cost-function
weighting volume. This weighting volume needs to be drawn by hand
in the partial-brain EPI space and should be set to 1 everywhere
except in the areas where distortion and signal loss are evident.
In these areas the weighting volume should be 0. If you then use
this weighting volume together with 6 dof and the search setting
above, starting with the output from Nudge (see the terminal where
you started Nudge to see where the outputs are) then it should work.
This is what you'd need to do using the GUI only, but it would be
better to use the command line, where you can use the transformations
saved by Nudge rather than have to work from the interpolated
volume output.
The command line you would need for this will be something like:
flirt -in partial_brain -ref anatomical -dof 6 -nosearch -init
nudge_matrix.mat -inweight partial_brain_weighting_volume -out
partial2anat -omat partial2anat.mat
The output (partial2anat) should be aligned with the anatomical
volume when viewed with FSLView and the matrix (partial2anat.mat) will
be the one that you would need to use to do any other transformations
from the partial-brain-space to the anatomical-space.
I hope this is clear.
All the best,
Mark
On 21 Sep 2007, at 03:25, Mark L wrote:
> Hello,
>
> Fair warning: I'm brand new at this, so this may be a very ignorant
> question.
>
> I'm trying to coregister a partial-brain functional volume (16 3-mm-
> wide
> slices covering the occipital/temporal cortex) to an anatomical
> volume, and
> FLIRT is having a hell of a time with it. I'm using the GUI, in the
> InputImage->ReferenceImage mode. I've tried 6, 9, and 12 dof
> models, and all
> three different search methods (Virtually aligned, same
> orientation, and not
> aligned), and most every combination of those. FLIRT has put the
> functional
> image everywhere but the occipital lobe.
>
> I'm sure I need to seed the model's fit somehow, i.e. tell it where
> to start
> - but I'm not sure how to go about doing that. In Brain Voyager
> (which I'm
> trying to leave behind), there's an "initial alignment" based on
> information
> in the functional volume's header files; is there something akin to
> that
> that I can do in FLIRT?
>
> Please help,
>
> !Mark
|