[Mrtrix-discussion] Maximum track length using streamtrack

Mon Jan 9 04:16:56 PST 2012

Hi everyone,

I didn't really see it as a "bug", because I was also already reasoning according to some of the logic explained by Donald here.  Actually, I'm personally also in favour of keeping it this way; otherwise a lot of other complexities/interpretation issues concerning the outcome of a fiber tracking and its meaning may indeed arise (based on how the tracking was done in the bidirectional case: one side first, or maybe both sides together at an "equal speed").  The way it is now, the outcome is quite clearly defined (which, I think, is of major importance in the interest of scientific use).
But I would indeed document it a bit better, so as to not confuse people.  From a naive point of view, it felt strange that I still got tracks when the "maximum track length" was smaller than the minimum track length (but no longer so when it was less than half of the minimum track length).  The "maximum track length's" definition is not (directly) comparable to the minimum track length's one.  Even that's not really a problem (according to me at least), as long as people who use it and report about it in their publications are well aware of it. :-)

Kind regards,
Thijs

From: Donald Tournier [mailto:d.tournier at brain.org.au] 
Sent: maandag 9 januari 2012 2:05
To: Robert Smith
Cc: Thijs Dhollander; mrtrix-discussion at public.nitrc.org
Subject: Re: [Mrtrix-discussion] Maximum track length using streamtrack

Hi all,

Just thought I'd add my 2 cents to the debate. The maximum length does indeed refer to the maximum distance from the seed point, and I agree that it could have been documented more explicitly. While I can see that it sounds like a bug to allow streamtrack to generate streamlines that are twice as long as specified, there is actually some logic behind it. The main purpose of that option was to terminate streamlines that had started to loop around themselves, or had done a U-turn, or taken some other such implausible path. This is why the default maximum length is 20cm - the length of your average brain. When using unidirectional tracking, this is fine, but things get a bit messier with bidirectional tracking. If you terminate the streamline in one direction because it's got too long, and enforce the maximum length strictly, you won't allow bidirectional tracking from that seed point. 

This may or may not be fine, depending on your point of view. The way I see things, the fact that the track might have got itself into a tangle on one side of the seed point should not mean that tracking should not proceed on the other side - the fact that it's got that long in the first place probably reflects a failure of the tracking to terminate appropriately, rather than a genuine anatomical feature, so whether you terminate at 20cm or 40cm is irrelevant. There is a good argument to be made for discarding the track altogether on those grounds, but if it's going to be kept, then I think the other direction should be tracked too. 

There is the associated problem that you might be using the maximum length for some other purpose, for example to look at the tracks within a certain distance of your seed region. In this case, you won't want to have your tracks discarded due to their having exceeded the length threshold, but I can see that in certain cases you might be surprised to find that many of them are longer than the length you specified. Hopefully making the documentation clearer will help...

Not sure what your take on it is, but we're open to suggestions as to how best to handle this option - bearing in mind that there are many ways to do tracking, and that the behaviour of that option should be at least vaguely sensible in all cases...

Cheers,

Donald.

On 9 January 2012 11:20, Robert Smith <r.smith at brain.org.au> wrote:
Thijs

Nice pickup; indeed yes, the tracking is allowing streamlines to continue up to the specified maximum length in either direction from the seed, such that it is possible for streamlines to be generated which are up to twice the 'maximum' length.

If you want to enforce your specified maximum length as a constraint upon the entire streamline rather than each unidirectional half-streamline, the easiest fix is as follows (in reference to the 0.2.9 code):
* in src/dwi/tractography/tracker/base.h, line 61:
bool set (const Point& seed, const Point& seed_dir = Point::Invalid) { pos = seed; num_points = 0; return (init_direction (seed_dir)); }

change to:
bool set (const Point& seed, const Point& seed_dir = Point::Invalid) { pos = seed; return (init_direction (seed_dir)); }

* in src/dwi/tractography/tracker/base.cpp, line 115:
void Base::new_seed (const Point& seed_dir)

At the start of this function (around line 117), insert this line:
num_points = 0; 

I will see about making this the default behavior for the next MRtrix release.

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        www.brain.org.au

On Fri, Jan 6, 2012 at 9:55 PM, Thijs Dhollander <thijs.dhollander at uzleuven.be> wrote:
Hi everyone,

Using Mrtrix 0.2.9, I noticed the following: when setting the -length parameter with streamtrack ("max_dist" as reported by track_info afterwards), the actual maximum track length seems to be twice that value.  I suspect each track is allowed to continue up to the "-length"-distance in both directions from the seed, i.e. allowing the full track to be maximally twice the "-length" value in size?

Happy 2012 & kind regards,

Thijs
_______________________________________________
Mrtrix-discussion mailing list
Mrtrix-discussion at www.nitrc.org
http://www.nitrc.org/mailman/listinfo/mrtrix-discussion

_______________________________________________
Mrtrix-discussion mailing list
Mrtrix-discussion at www.nitrc.org
http://www.nitrc.org/mailman/listinfo/mrtrix-discussion

-- 
Jacques-Donald Tournier (PhD)
Brain Research Institute, Melbourne, Australia
Tel: +61 (0)3 9035 7033