If you use the seeds to targets classification approach, you will get an
image that says how many samples from each seed voxel reached each target.
If you just use a single classification target this won't tell you a whole
lot beyond which voxels in your BA10 mask connect to BA40. Probably most or
all voxels will have a few samples connecting, so you will need to threshold
this somewhere. What this method will not tell you is the path the samples
take to get to BA40 (which seems like what you are interested in).
If you want to avoid a priori assumptions on the route the path takes to
BA40, simply do a single ROI tractography from BA10. Then you can select
those paths that go to BA40 in a second step with a waypoint mask you draw
around the BA40 terminations from the original result. I use this kind of
multi-step approach all the time when defining fascicles like the arcuate.
One other thing to keep in mind is your tractography result from BA10 to
BA40 will likely not be the same as the one from BA40 to BA10. Thus it is a
good idea to track both ways and sum the result (symmetric tracking). The
reason for this asymmetry is the diffusion orientation field one encounters
going in one direction is not the same as the diffusion orientation field
one encounters going the other direction. Many branch points tend to only
be acceptable going one way (think of the difference in angle of the turn
you would have to make when approaching a highway on ramp from the correct
vs the incorrect direction).
The length of time it takes to run tractography depends mainly on the number
of seed voxels and the number of samples you send out from each. Adding
waypoint masks or classification masks doesn't change things much. If you
are using cortical seeds, you can speed things up considerably if you just
use seeds along the white matter surface (1 voxel thick).
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Torsten Ruest
Sent: Thursday, June 10, 2010 12:19 PM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx options
Hi Matt,
I am doing the classification with both hemispheres, independently however
as I thought it would change the fdt_paths file if I use 2 classification
masks. I am pretty much new to this, so I am playing around with all the
options - I do the waypoint option in a separate process too, but these
things take a while to run. I would have thought that in the approach with
the 2 independent classification masks, the seed_* images would be identical
to the ones I'd obtain from using 2 classification masks. No?
Well I hoped the software would tell me which voxels from BA10 connect to
BA40 without hypothesis driven waypoint masks. Ultimately, I'd like to match
up our (to be acquired) fMRI data with the tractography - here we expect a
facilitated deactivation of BA40 due to drug in a task condition and an
increase in BA10 activation also due to drug compared to non-drug (I'd
possibly use the activation clusters then as masks...). So I'd like to
explore if there are any potential tracts to have a closer look at, as we
also look for genotype interactions that could possibly facilitate drug
response even more and these interactions may have their origin in altered
white matter integrity.
Thanks,
Torsten
|