[Mrtrix-discussion] ACT: minimum streamline length criterion

[Robert Smith]

Hi Ulrike (and all),

After submitting that manuscript and experimenting more with the raw data,
I ended up changing my mind about the whole white matter fraction path
integral calculation in ACT.
The theory was to allow very short streamlines that confidently passed
through white matter, but require a longer length from streamlines that
consistently pass through areas with partial volume with other tissues... I
liked the concept, but in practise it introduced some issues:

   - Where there was an absence of short reconstructed streamlines, it was
   difficult to tell whether streamline rejection was caused by this criterion
   or some other mechanism;
   - Without densely up-sampling the streamlines (which would be costly),
   the result of the path integral calculation could vary depending on exactly
   where the streamline points lay within the sub-voxel volume;
   - Without a clear tangible benefit, it was kind of disobeying Occam's
   Razor.

The current default tckgen behaviour is as follows: The minimum length is
automatically set to 5 voxels, *unless* ACT is used, in which case it
defaults to 2 voxels; either of these is over-ridden using the -minlength
option. The white matter fraction path integral calculation is disabled by
default. There's a pseudo-justification for the 2-voxel-length threshold in
the SIFT follow-up paper
<http://www.sciencedirect.com/science/article/pii/S1053811914008155>.

If you want to play with it, the path integral method is still there in the
code, it's just not accessible from the command-line. You can change the
values by directly modifying file src/dwi/tractography/ACT/act.h, lines 33
& 34, then recompiling: The first is the white matter fraction path
integral threshold, the second is a threshold applied to the maximum WM
partial volume fraction along the streamline (bear in mind that it's a
trilinear-interpolated value read at each point). The standard minimum
length criterion will be applied also unless you specify -minlength 0.

Would like to hear about any tests you perform, or whether anybody has
particularly strong views on the topic.
Rob


--

*Robert Smith, Ph.D*
Research Officer, Imaging Division

The Florey Institute of Neuroscience and Mental Health
Melbourne Brain Centre - Austin Campus
245 Burgundy Street
Heidelberg Vic 3084
Ph: +61 3 9035 7128
Fax: +61 3 9035 7301
www.florey.edu.au

On Mon, Dec 22, 2014 at 10:00 PM, Ulrike Kuhl <kuhl at cbs.mpg.de> wrote:

> Dear MRTrix team,
>
> I want ot use anatomically constrained tractography as explained in the
> Smith (2012) paper using the 'tckgen -act ...' command.
> However, I'm a little unsure about the minimum streamline length option:
> if I do not call 'tckgen -act ...' with a specified '-minlength' option,
> does the algorithm automatically use the white matter fraction path
> integral criterion as explained in Smith (2012)?
>
> Thank you very much for your help!
> All the best and have a happy Christmas!
>
> Ulrike
> _______________________________________________
> 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/20141223/29568876/attachment.html>