Print

Print
Hi,

>it seems as though the Bayesian implementation of probtrackx is most accurate

even though the text and Fig. 19.13 may suggest that, I would not generally say 'most accurate' in a strict biological sense. I would probably say >the constrained Bayesian approach was less prone to false-negatives than the streamlining algorithms examined, and it tends to be more specific than 'ordinary' bedpost/bedpostx<.
The reason is that pyramidal tract volumes tend to get overestimated by bedpost/probtrack(x) for the tracking between the handknob and the cerebral peduncle when no other waypoints/exclusion/termination masks are used. Furthermore, presumed tract (volume) symmetry between the hemispheres may be misleading in brain tumors with perifocal edema.
However, all this depends on how you do the tracking (e.g. you can increase the 'specificity' of your bedpost/probtrack(x) results by choosing anatomically sensible waypoints/exclusion/termination masks and other 'streamlining' algorithms may outperform the ones I specifically tested) and what you track (e.g. it may be different for other tracts which are crossed by more fibers such as the IFO or ILF) so the generalizablitity is somewhat limited.
hth,
Andreas




Von: Andy Alexander <[log in to unmask]>
Antworten an: FSL - FMRIB's Software Library <[log in to unmask]>
Datum: Mon, 9 Jan 2012 12:34:52 -0600
An: <[log in to unmask]>
Betreff: Re: [FSL] bayesian tractography

Hi Stefano,

I believe that Freesurfer has a Bayseian implementation of probtrackx in its latest release.    One issue for surgical purposes is that the Bayseian approach assumes that the anatomy is consistent with their training set.  I don't think it would work very well if you have significant mass effect or deviant anatomy.

Best,

Andy

On Jan 9, 2012, at 12:24 PM, Marenco, Stefano (NIH/NIMH) [E] wrote:

In reading Andreas Bartsch’s book chapter on different implementations of tractography to the problem of tract definition for surgical purposes, it seems as though the Bayesian implementation of probtrackx is most accurate. Is this already implemented in the current FSL version or is it something to come (and if so how far down the line)?

 

Stefano Marenco, MD

NIMH/GCAP

Building 10, room 3C104

10 Center Drive

Bethesda, MD 20892

tel 301 435-8964

fax 301 480-7795

email: [log in to unmask]

 


Andrew Alexander, Ph.D.
Waisman Laboratory for Brain Imaging
Departments of Medical Physics and Psychiatry
(608)265-8233

Waisman Center
1500 Highland Ave
Madison, WI 53705