If you only use only the thalamic seed and
cortical target in one hemisphere, I would be surprised if you got much in the
other hemisphere. If you still have that issue, even when using the white
matter mask below, you could put a single slice mask across the corpus callosum
and call it with the –avoid argument (I think this is exclusion in the
GUI?).
Peace,
Matt.
From:
Sent: Wednesday, March 25, 2009
1:45 PM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx and
thalamocortical tracts
I am trying to detect only the tract
between the thalamus and the PFC. I will run the program again as a single seed
mask (thalamus) and the PFC being the cortex.
The other masks I would like to use are
termination or exclusion masks to constrain further the fibertracking (e.g. to
one hemisphere). Stefano
From: Matt Glasser
[mailto:[log in to unmask]]
Sent: Wednesday, March 25, 2009
1:04 PM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx and
thalamocortical tracts
I think you don’t want multiple
masks in this case, you just want the thalamic seed and cortical target (or
waypoint) (this is based on what you said in the first e-mail). Now you
are talking about three masks, and I’m unclear what it is you are trying
to do. Can you explain exactly what tracts you are hoping to find, and
how they will pass through the masks you are using?
Peace,
Matt.
From:
Sent: Wednesday, March 25, 2009
11:52 AM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx and
thalamocortical tracts
Hi Matt and other tractography experts, I
like your suggestion for the white matter mask, and maybe the easiest way to
resolve both of my questions is to cut that mask further behind and below the
thalamus to avoid stray samples from continuing into unwanted territory.
However, I am going to ask a few more questions about doing the tractography
below.
I think your other suggestion of seed in
thalamus and target in ctx does not work, as the tract extends behind the
thalamus, thus further termination masks may be necessary. I did this by using
the gui and selecting multiple masks, listing the thalamic and frontal cortex
masks as seeds and the prefrontal cortex as termination mask. I would not use
exclusion masks because those would eliminate the whole length of the sample
extending from thalamus to dlpfc and propagating behind and below the thalamus.
Is this last statement correct?
Another question I have is whether by
specifying two masks, probtrackx does tractography using both of these
independently as seeds, but maybe no tracts will emerge from the prefrontal
cortex mask because it is also specified as a termination mask.
One other problem is that only one
termination mask is allowed in multiple masks tractography, at least in the
gui. Would it be possible to use further termination masks when using the
command line?
The normalization remains an issue, as
with multiple masks tractography the waytotal is not created, I think (at least
I don’t see it in my output dir)…. So the only way is to use the
first solution. In the case that I mentioned above, the max value I get is
18329, somewhere in the thalamus. I would expect this voxel to have a fairly
high probability of connection after normalization (e.g. 0.9), but given 280
voxels in the thalamic mask, 6309 in the PFC mask and 5000 samples per voxel,
it seems that this would not be the case at all. So I am sure I am doing
something wrong, I am just not sure what or how I should prescribe the
tractography. Any help is appreciated.
Stefano Marenco
From: Matt Glasser
[mailto:[log in to unmask]]
Sent: Monday, March 23, 2009 4:51
PM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx and
thalamocortical tracts
Well regarding your second question, you
could segment the white matter using your structural image using FAST or other segmentation
method, make a mask containing only the white matter, transform it to diffusion
space using FLIRT and –interp nearestneighbour, and use that instead of
nodif_brain_mask. I do this all the time (with segmentations generated
from FreeSurfer white matter surfaces) and I get much cleaner tractography
results as a consequence. If you put your seed in the thalamus and your
target in the cortex, you should avoid pathways going beyond the
thalamus. As far as normalizing your probability values, you can either
divide by the total number of samples you send out (# of samples per voxel X
number of voxels) or by the number of samples that reach the target (the
waytotal), depending on what you want to show. I have found that the
first method shows hemispheric asymmetries better, whereas the second method
gives a more consistent pathway size across hemispheres and subjects.
Peace,
Matt.
From:
Sent: Monday, March 23, 2009 3:40
PM
To: [log in to unmask]
Subject: [FSL] probtrackx and
thalamocortical tracts
I have two masks: one in the thalamus and the other in the
cortex and I want to identify the tract that connects these two masks, nothing
more.
Is there a way to do this without creating termination or
exclusion masks below and behind the thalamus?
Is there a way to do this that would avoid assigning very
low but non-zero values to the gray matter and /or CSF near the white matter?
In the case where I would need to threshold the resulting
probabilities in order to exclude gray matter or CSF (say pick voxels with 90%
probability of belonging to the tract), if I were running probtrackx in
multiple masks mode, would it be sufficient to divide by the resulting image by
the total number of voxels in the two seedmasks to obtain reasonable
probability values?
NIMH/GCAP
Building 10, room 3C104
tel 301 435-8964
fax 301 480-7795