Dear Hongyu,
I'm uncertain as to whether you have a registration problem
or whether you simply have a large brain mask and some
activations in the mask which are technically outside the brain.
This is not uncommon due to ghosting and other acquisition-related
issues, and is no real problem. If you look at your first level
results,
which are shown on the original functional images, then you can
figure out whether you have activations outside of the brain like
this or not. If you do, then your problem has nothing to do with
registration, as the first level results do everything on the original
functional images and do not use any registration results. However,
if you do not see any activations outside of the brain in the first
level results but then see them at higher levels then this is likely to
be a registration-related issue.
Can you let us know which it is?
All the best,
Mark
On 4 Mar 2010, at 20:22, Hongyu Yang wrote:
> Hi Mark,
>
> We checked all the registration page results, based on our limited
> experience, they were OK. We even compared the results with the
> previous
> results we got before, in which the activated areas of post-stats
> were
> matched with the background brain image, still could not find some
> obvious
> differences between the two. I tried to use some introduction from
> FLIRT,
> which said If we want to register fMRI data with only a few slices,
> following the instructions of this link:
>
> http://www.fmrib.ox.ac.uk/fsl/flirt/ztrans.html
>
> However, I still got similar results, activated areas beyond the
> background
> brain image. One of my colleague said he don't think it's a
> registration
> problem. He think we have noisy voxels outside brain that happen to
> follow
> the model, and then don't get masked out because the brain mask is a
> little
> bigger than brain. I can not agree with his points, however, I do
> not know
> what problem caused this consequence. Can I send some results to
> you, and
> can you help us to come out from the dilemma, which bothers us for a
> very
> long time.
>
>
>
>
|