To get better brain extractions on 3T images, I use the following
* pre-mask the image with standard_space_roi -b
* bias-correct the output with fast
* bet on the _restore image with -R and -f 0.4
Most of the time, this gives me a mask without false negatives and an
acceptable number of false-positive voxels. You could conceivably
mask the original with the resulting bet mask (use -m to get it) and
feed the result to fast again, but so far I haven't found this to be
Annoyingly, bet still tries to determine the vertical axis from the
data ordering rather than the header information, so if the data
ordering is different from what bet expects, it can be useful to run
fslswapdim RL PA IS on your input.
Hope that helps
On Thu, Jul 23, 2009 at 02:11:32PM +0100, Jay Ives wrote:
> I am intending to do VBM, but have noticed BET is not including areas of
> cortex in some places, particularily at the vertex and frontal poles. This is
> clearly going to be unacceptable.
> I have tried various alterations to parameters without success. BET does not
> give me this problem on images from 1.5T.
> Could this be due to some bias field in my 3T images? They do look extra
> bright in the centre of the images.
> If so, is there some tool or process to even out the intensity prior to applying
Rolf A Heckemann, PhD
Division of Neuroscience and Mental Health
MRC Clinical Sciences Centre
Imperial College London
Hammersmith Hospital Campus
Du Cane Road
London W12 0HS