I'm having issues getting X11 to show me all the flags for fslmaths and fslstats (my scroll bar isn't working and I can't figure out how to have it do one page at a time), so I'm not sure if these are the best flags.  Although, from what I can see, it looks like the -k flag in fslstats is the one I want (help says "use the specified image (filename) for masking - overrides lower and upper thresholds").  Does this look right (also getting the mean FA and volumes)?


fslstats Normed&Thresholded_fdt_paths -k FreesurferWMsegmentation -M -V


-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Matt Glasser
Sent: Sat 3/28/2009 9:18 PM
To: [log in to unmask]
Subject: Re: [FSL] Post probtrackx analyses - thresholds for fdt_paths
 
FreeSurfer certainly has a segmentation that you can use, since I don't use
the segmentation directly from FreeSurfer, rather I generate it in Caret
from the FreeSurfer white matter surface, I don't know the exact file you
should use.  You can simply mask the tractography result with the FreeSurfer
segmentation in fslmaths (or even in fslstats).

Peace,

Matt.

-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Lindgren, Kristen, Ann
Sent: Saturday, March 28, 2009 7:20 PM
To: [log in to unmask]
Subject: Re: [FSL] Post probtrackx analyses - thresholds for fdt_paths

Matt, quick follow-up question.  You had mentioned in a previous e-mail that
instead of thresholding FA it would be more appropriate for me to confine my
thresholded masks within a white matter segmentation.  I was thinking that I
could use the white matter segmentation from freesurfer (one for each
hemisphere) for this.  Any ideas on how I could do this?  Is it something I
could do within fslmaths or is it something I'd need to define when I run
probtrackx?  Thanks!

Kristen


-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Matt Glasser
Sent: Sat 3/28/2009 7:48 PM
To: [log in to unmask]
Subject: Re: [FSL] Post probtrackx analyses - thresholds for fdt_paths
 
Well, I draw the ROI on the cortical surface in Caret and then map it to the
volume, which makes a 1 voxel thick ROI in the voxels of the white matter
surface.  There are a number of ways to generate the white matter surface,
however.  Before I started using freesurfer, I used the FAST segmentations.
I took the white matter mask from FAST, blurred it with a 3x3x3 voxel
kernel, and then selected voxels with values between .50 and .85.  This
produced a white matter surface ribbon that is about 1 voxel thick.  You
could then find the intersection between your ROI and the ribbon to get the
cortical ROI that is only in the first white matter voxel.  

 

I always correct for distortions (there are a variety of ways to do this, a
phase up/phase down correction is best, but you can also use a field map or
even use FNIRT to register your FA to your T1 and then apply the warpfield
to your raw data).  The issues I describe below occurred from tractography
from ROIs in diffusion space.  It is worse in some brains, or acquisitions
than others, but I just choose to avoid it across all subjects, rather than
make exclusion/termination masks on a case by case basis.  I am not
convinced that you get more information by tracking into the cortical grey
matter, given the resolution of DTI, than you get from tracking up to the
superficial white matter, and you can avoid errors this way.

 

Peace,

 

Matt.

 

  _____  

From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Martin Kavec
Sent: Saturday, March 28, 2009 7:36 AM
To: [log in to unmask]
Subject: Re: [FSL] Post probtrackx analyses - thresholds for fdt_paths

 

Hi Matt,

On Fri, Mar 27, 2009 at 10:54 PM, Matt Glasser <[log in to unmask]> wrote:

I always track from white matter.  If it is a cortical ROI I use the 1st
voxel of the white matter surface.


That's an interesting approach. How do you get from cortical ROI to the 1st
voxel of the WM?
 

 Honestly, I have found that tracking
from/into cortical grey matter results in more error than useful information
(its easy to get pathways crossing sulci for example, if you combine partial
voluming with the fact that the principle diffusion direction in cortical
grey matter is oriented perpendicular to the cortical surface, and cortical
grey matter ROIs may be very close to brain rim voxels).


Could this simply be because of the nonlinear distortions between DTI and
FreeSurfer input image? It could be several voxels in most of the brain.

Thanks,

 Martin