A 2 voxel border would certainly be fine - probably even 1 voxel.
As for inverting, just do the following:
fslmaths originalmask -mul -1 -add 1 -thr 0.5 -bin invertedmask
All the best,
On 21 Aug 2009, at 23:00, Anastasia Ford wrote:
> Thank you both for your replies. I will try the registration
> following your suggestions. If I
> have a 2mm*2mm*2mm DWI data, would including a 2 voxel border around
> the lesion in
> addition to the lesion mask itself be a good choice? Also, could
> you suggest how to invert
> the lesion mask so that the in-lesion voxels have a value of 0
> whereas the rest of the brain
> has a value of 1? Thank you.