Hi Joe,
On 7 Mar 2008, at 16:01, Joseph T. Devlin wrote:
> Hi folks,
>
> My apologies for a really basic set of questions but I discovered
> that I've never thought much about the basis of the brain mask
> before and now it's tripping me up...
>
> If I understand correctly, in first level analyses the brain mask is
> essentially generated by first motion correcting the 4D data, then
> running bet on it (using the -F option), and then thresholding the
> volumes at 10% and keeping only those voxels which were non-zero
> across the time series.
Yes - it's just supposed to be a mostly-unimportant, liberal brain
masking. It shouldn't matter if it is over-liberal - I've not seen any
cases where that was a problem.
If the 10% intensity thresholding is too high (it isn't normally?)
then you can just lower that under the 'misc' tab, even down to 0 if
necessary. Hopefully the voxels that you're losing are just down to
the thresholding.
> When I do this with our data (1.5T Siemens, GE-EPI -- fairly
> standard stuff), the mask has two problems:
> 1) it leaves in all the scalp and stuff surrounding the brain and
> 2) it leaves out individual voxels inside the brain.
>
> In the first case, this seems to come from the fact that the bet .. -
> F command uses a lenient idea of brain and then further dilates the
> mask, which essentially includes the splodge surrounding the brain
> (but it does a good job removing eye balls). I wondered why the
> extra dilation step was included and whether it can turned off or
> modified?
You can always tune the call to BET but I don't think it's a problem
to be over-inclusive at this stage? This doesn't affect the
registration as example_func has already been created.
> For the second, the fact that -Tmin is used in the command to
> generate the mask significantly increases the number of missing
> voxels within the brain volume due to small head motions interacting
> with areas of low intensity (ie small spaces such as the lateral
> horns of the ventricles near the hippocampus or places where noise
> in the data artificially reduced the signal at a single time
> point). I was sure what the motivation for including the -Tmin
> option was -- is it just to reduce the volume to 3D or is it serving
> a deeper purpose?
It's to exclude voxels which are dodgy to use for exactly this
reason ;-) Do you really want to include such voxels who's motion is
taking them in-and-out of being 'valid'? You can always lower the
threshold though :)
Hope this makes sense? Happy to iterate more on this if needed.
Cheers, Steve.
> Finally, I wondered why the mask was generated in the current
> fashion rather than simply using bet to strip out the rubbish (I
> mean, of course, the essential non-brain material...) in
> example_func.nii.gz? Ok, and finally for real -- whether there was
> a method for ensuring that individual voxels within the brain were
> not excluded -- that is, computing the surface and then accepting
> everything within that volume as brain material? At the moment, CSF
> is not masked out because it has a higher intensity in T2* images so
> I can't see any reason not to just assume anything within the
> surface is wanted.
>
> I assume there are good reasons for these choices but I realized I
> don't have any good idea what they might be.
>
> Thanks, in advance, for your help!
> Joe
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|