You would typically draw the ROIs in FSLView, and then save them. The
simiplest thing to get working is to draw your ROIs on one of the diffusion
scans (e.g. the FA) and then track from that. If you want to draw your ROIs
in a different space, like on the subject's structural image or in a
standard space, like the FMRIB58_FA_1mm (MNI space), you will need to
calculate a transform between that space and your subject's diffusion space
by registering an image in one space to an image in another space. You
might start by registering your subject's FA image to the FMRIB58_FA_1mm
using FLIRT. A more advanced and somewhat more accurate registration method
would be to first do linear registration with FLIRT and then nonlinear
registration with FNIRT. The tracking program, probtrackx, cannot yet
accept the warpfields output by FNIRT, so you would need to do the transform
of your ROI using applywarp (you can do both linear and nonlinear components
at the same time). Be sure to use nearestneighbour interpolation. For only
linear transforms, you can simply specify the transform in probtrackx --xfm.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of RUSSELL JARVIS
Sent: Monday, May 25, 2009 6:39 PM
To: [log in to unmask]
Subject: [FSL] Creating masks for target regions in DTI Probtrack
Hi everyone,
I wonder if anyone knows how to create masks for optional targets in FSL DTI
prob track?
I am thinking it involves using a brain atlas in FSL view, but I am unsure
if I will have to manually draw slices using the lightbox view, or if I can
automatically turn a selected atlas region into mask?
If you can automatically turn a selected atlas region into a mask, does
anyone know about the steps required to do this? I have a standard image:
FMRIB58_FA_1mm loaded with the Juelich Histological Atlas on, and I have
opened the MNI structural atlas and selected the frontal lobes. I have 5
seed masks that correspond to regions of the corpus callosum, and I would
like to get the probabilities associated with fibres from each submask
reaching some targets regions ie frontal lobes etc.
If you have to manually draw over selected regions it could be very
difficult to create masks as I think this would involve drawing surfaces
rather than planes. it almost seems unachievable to get a reasonable level
of accuracy by eye.
Thanks.
From Russell.
|