Hi, Mrtrix experts!
I’m trying to make poster with tracts from Mrtrix3. I’ve used HCP data and pipeline from this page: https://mrtrix.readthedocs.io/en/latest/quantitative_structural_connectivity/ismrm_hcp_tutorial.html?highlight=MSMT-CSD
The only settings I changed:
tckgen … -select 10M -cutoff 0.05
tcksift … -remove_untracked
On the screenshot you can see inferior longitudinal fasciculus tract from mrview.
The fibers looks correct, but its tortuosity is too high. This makes bad impression from the image - the tracts looks like ramen noodles.
Can you recommend me some tracking or mrview settings for making smooth fibers?
Hi Eduard,
If you want smooth tracts, is there any reason you haven’t used deterministic tractography? You can also play with the -power or -cutoff flags to get “smoother” tracts but they are going to slightly change the interpretation what the streamlines represent (as would deterministic tracking). If you simply want to extract well known tracts, I do not really think that this will be an issue.
Claude
The default tracking algorithm (iFOD2) uses much more conservative estimates of fibre orientation uncertainty than other diffusion models / probabilistic streamlines algorithms, which is what leads to the “noodly” nature of the streamlines. The -power option will directly influence the magnitude of this effect, with a power of infinity theoretically turning iFOD2 into a deterministic algorithm. There have been multiple long-winded discussions and experiments with @jdtournier trying to figure out what this parameter “should be”, but unfortunately we’ve not yet come up with a single unambiguous answer to this. There are some precedents in the literature for increasing this parameter in order to make the results of tractography look “nicer”, but we can’t use that logic to increase the default value of this parameter for all applications, as it may lead to false negatives.
It is theoretically possible to post-hoc smooth streamline data, but we don’t currently have code for doing so.
Rob
@cbajada, @rsmith, thank you for answers!
Deterministic algo SD_STREAM produced just a few tracts. Such image is not appropriate.
I played with -power option in iFOD2 algo. Tckgen with -power 10 yields good result.
1 Like
