Hi Stefan (and Markus),
In probabilistic tractography, what you calculate is a "spatial"
histogram that tells you the chances of passing through any location.
The histogram is calculated by sampling trajectories, which do not
have "theoretical" stopping criteria. You may reject/interrupt a
sample trajectory on the basis of some prior anatomical criteria
(reaches the edge of the brain, loops on itself, or else).
In your case, if you want to calculate the proportion of samples that
stop somewhere, you need to define what you mean exactly by "stop
somewhere", among the following options:
(i) a sample reaches a termination mask
(ii) a sample reaches the edge of the brain
(iii) a sample loops on itself
(iv) a sample reaches a location where the next sample orientation is
deviated by an angle higher than some threshold value (curvature
threshold)
So it doesn't seem obvious to me how useful a measure of the
proportion of stopping samples could be.
However, you may certainly calculate the probability that A connects
to B via C vs the probability that A connects to B without passing
through C. This seems to be a useful measure for the cases you've been
describing, no?
Cheers,
Saad.
On 1 Dec 2008, at 13:55, Stefan Kreisel wrote:
> Hi Markus,
>
> your suggestion is completely plausile. However, consider the case
> of large irregular masks. One could dilate a given waypoint by say
> one voxel in 3D and use this as the termination mask -> here,
> however, one has no real way of defining what's proximal and what's
> distal. This is the exact problem that led me to ask if there isn't
> some set theoretical way of defining probability of termination...
>
> Regards,
>
> Stefan
>
Saad Jbabdi
Oxford University FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
+44 (0) 1865 222545 (fax 717)
www.fmrib.ox.ac.uk/~saad
|