[Mrtrix-discussion] seeds and whole brain tractography

[Robert Smith]

Hi Ertan

This has been the major subject of my PhD research, so I'll try to give you
a coherent response without boring you with technical details or
philosophical ramblings.

Although it seems a sensible thing to do, this reduction of tractography
streamlines is technically challenging, and researchers have been trying to
come up with an appropriate method for the last 10 years (a literature
search for 'tractography clustering' should give you plenty of hits). It
may seem trivial to label a pair of streamlines as 'similar' by somehow
comparing their trajectories, but it is deceptively difficult to use this
information to segment the streamlines data in a manner which is robust,
reliable and in any way meaningful. There is also a scaling problem in
dealing with the large number of streamlines that you need to achieve an
adequate reconstruction of the white matter of the whole brain.

Depending on future developments, there is a chance that my more recent
method (abstract 673 here <http://www.ismrm.org/11/Session67.htm>,
apologies for the shameless self-reference) may be included in a distant
version of MRtrix. If however you are using deterministic streamlines, and
are looking to reduce the 'redundancy' associated with multiple
deterministic streamlines reconstructing near-identical pathways, a method
such as that by Guevara et
al.<http://www.sciencedirect.com/science/article/pii/S1053811910013200>may
be more appropriate. However I am not aware of any tractography
clustering method which has been made available to the public thus far
(probably because developers such as myself are not confident enough in
their performance to release them).

So although the answer may be 'no', hopefully this provides some food for
thought.

Regards
Rob

--

Robert Smith
Melbourne Brain Centre
245 Burgundy Street
Heidelberg VIC 3084
Telephone: (+61 3) 9035 7128
Fax: (+61 3) 9035 7301
Email:  r.smith at brain.org.au
www.florey.edu.au <http://www.fni.edu.au>        www.brain.org.au



On Wed, Dec 7, 2011 at 9:43 AM, Ertan Cetingul <ecetingul at yahoo.com> wrote:

> Hello,
>
> I have a question on deterministic whole-brain tractography. Given a "seed
> mask" (e.g. FA-based thresholded mask), I presume that the algorithm
> considers each voxel in the mask as a starting seed point and performs
> streamlining. Now consider the following hypothetical scenario: Suppose we
> have 100 voxels (1x100 formation in 1D) with tensors indicating a linear
> fiber from left to right. If the algorithm is repeated at every voxels,
> theoretically one should obtain 100 copies of the same fiber. In reality,
> we are expected to find 100 similar fibers (in terms of spatial positions,
> length, etc.). I think one should not say that the remaining 99 copies are
> spurious but I believe there is a need to resolve this "dense"
> representation of the same structure. I was wondering whether the MRtrix
> contains a seed selection strategy to resolve this issue (and speed up the
> process) or a fiber similarity-driven filtering/pruning strategy after
> tractograpy.
>
> Thanks
>
> Ertan
> _______________________________________________
> Mrtrix-discussion mailing list
> Mrtrix-discussion at www.nitrc.org
> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20111207/c406442c/attachment.html