Hi,
What inputs/options are you giving BET? The main thing you might want to
play around with is the fractional intensity threshold (-f), depending on if
the extraction is leaving too much or too little non-brain matter behind.
But, you might want to adjust some of the other defaults as well. I don't
think you should have to blur the image to achieve good results.
I normally only have to lower the -f to about 0.35 to get good results on my
anatomical images.
bet ${SUBJ}_anat ${SUBJ}_anat_brain -f 0.35
One final thing to note is bet generally works best on images that do not
have a lot of excess non-brain matter (e.g., neck tissue). However, I
believe bet2 can deal with this if you implement the -R option...
http://www.fmrib.ox.ac.uk/fsl/bet2/index.html
Hope this helps,
David
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Dav Clark
Sent: Tuesday, March 03, 2009 5:43 PM
To: [log in to unmask]
Subject: [FSL] Getting good extractions
Hi all,
I'm struggling currently with FSL and getting proper brain extractions
with BET. It seems that everything requires BET be run first... so
there's no way inside of FSL to get a good result if BET doesn't
work. The only idea I've seen is that you can smooth a structural
heavily, then presumable subtract that image out from your base image.
That seems a little troubling... wouldn't that also get rid of some
contrast?
DC
|