Hi,
On 25 Oct 2007, at 14:45, David V. Smith wrote:
> Hi-
>
> I'm using FSL 4.0.1, and I'm having some masking problems with FEAT
> once I get
> to the 3rd level analyses. The mask.nii.gz image created by FEAT
> masks out
> much more of the brain than I would like it to... It's essentially
> cutting off
> many parts of the brain and my stat maps look 'speckled' because of
> the mask
> (e.g., a big blob of activation can have a big chunk missing where
> you might
> expect to see a local maximum). This is particularly problematic
> in FFA and
> OFC. See attached images of 3rdlevel mask.nii.gz (shown in red)
> overlayed onto
> the 3rdlevel mean_func.nii.gz and also on the bg_image.nii.gz.
>
> 1.) Is this because these particular voxels are missing in some
> subjects (and
> shouldn't be used in the analysis) because of BET being a little
> too liberal at
> earlier stages in the analysis? I realized it's supposed to fairly
> conservative
> on the functional data, but clearly parts of my brains are getting
> cut off...
Yes. When you ran the second-level analysis there would be a
registration evaluation summary page, which you should look at
carefully, including all the registrations and the summary mask
images at the bottom. You should definitely check which of the
subjects were problematic.
You're right that the thresholding and BET settings at first level
should be overinclusive, so we don't often see this kind of problem.
If your data has some weird intensity range, for example, you might
need to change the %threshold setting (in the Misc tab) in the first-
level analyses.
This may fix things, also, if BET really isn't working well (unlikely
for FMRI data) then you can just turn it off in the Pre-stats tab.
As a last resort, yes you can over-ride the masking with a non-GUI
option - copy
mkdir -p ~/.fslconf
cp $FSLDIR/etc/fslconf/feat.tcl ~/.fslconf
and then edit ~/.fslconf/feat.tcl, changing the following line to
include a full pathname to the mask you want:
set fmri(alternative_mask) ""
Note that this probably only has effect at the first-level analyses -
there's nothing that you can do to "fix" the masks at higher-level -
there's no point trying to fix the masks as the input data (copes
etc) are already masked at first level.
Cheers.
> 2.) If it is related to the BET settings that FEAT/MELODIC use
> during pre-stats,
> how can I change these to something more appropriate (e.g., reduce the
> fractional intensity option)? If this is the fix, would I need to
> re-run my
> entire analysis?
>
> 3.) Is there a way to simply give FEAT a different mask instead of
> the one it
> generates? This seems to be the easiest fix and I thought one of
> the new
> non-GUI options (below) in the higher-level design.fsf would help
> me, but it
> didn't get used as the mask.nii.gz image like I expected. Is this
> not what
> this option is for?
>
> # Alternative (to BETting) mask image
> set fmri(alternative_mask) "insert_better_mask_here"
>
> Thanks,
> David
> <bg_image.nii.png>
> <mean_func.nii.png>
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|