Hi
Currently, I'm working on optimize FLIRT in order to better reslice DWI.
In order to emphasize a particular ROI, I've first masked the input volumes
of FLIRT (given with -in and -ref). As results were not convincing, I've
surch for more help and found this Web page :
http://www.fmrib.ox.ac.uk/fsl/flirt/overview.html
It's explained that directly masking input volumes introduce "artificial
boundaries" and I could verify that.
So, as suggested in the Web page, I've tried the -inweight and -refweight
option.
Result was definitely inconsistent (very big translation and rotation), even
with simple binary mask on the brain obtained after beting the input and
reference volumes.
I use FLIRT as this :
flirt -ref ref.nii -in in.nii -dof 12 -cost mutualinfo -refweight
refweight.nii -inweight inweight.nii
What values is suggested to weighted volume ?
Does 0 is accepted ?
When we use this option, is it better to use both otpion -refweight and
-inweight or only one can be used ?
Thanks for your help.
Franck
|