Hey, completely new to tractography, so not quite familiar on what the right output should look like. I ran through the HCP tutorial with the exact parameters given (btw the tutorial was great), and I got a weight matrix connectome. Some of the weights are 1 and some are like 600,000. I am very confused about this, does that mean the 1 weight is something like noise? Thanks
Yes, the variation in connection density of different edges in the connectome will vary massively. However this should not be written off as “noise”: There is no robust mechanism by which to separate spurious false positive connections from real low-density connections (on a per-subject basis at least). I have on multiple occasions cited the work of Nikola Markov, who reports quantitative connection densities between GM areas from tract tracing that vary by over five orders of magnitude; that is to say, in the brain there are GM areas that have huge numbers of fibres between them, and others that have very few (even one or two) axons between them. Therefore “density” is not conceptually an adequate number of “false-ness”.
The stochastic nature of the default tracking algorithm in MRtrix3 also contributes to this “spread” of connection densities: it is inevitable that with such a huge number of streamlines generated, and a comparatively high degree of orientation dispersion with respect to other software packages, some “low-probability” pathways will be reconstructed.
I would also suggest looking at this related article written by some collaborators of mine, if you are interested in how others have approached this observation.