MRtrix 3 release


Dear MRtrix Team,

we are planning to start a very large analysis soon.
Therefore we need to know if we can wait for the final release of MRtrix 3 or if we should use release candidate 2.

Will there be a third release candidate?
Do you have some sort of road-map for the release?
Do you expect changes in msmt reconstruction, fod2fixel or tckgen?

Thank you very much!


Yes, we are currently in the final stages of preparing the release of another release candidate, 3.0_RC3. This is to fix a bug that affects tckgen, which is likely to have an impact on the results. We are still investigating how big that impact might be, so we can provide adequate guidance to users (based on what we’ve seen so far, the impact is likely to be very minor, but not negligible). But I’d definitely hang on till then. I’m frantically trying to finalise the plans for the workshop in June (we want to open up for registration next month), so this issue has not been getting as much attention as I’d have liked – this would otherwise have been released by now.

In terms of road-map, we intend to release properly at some point before the ISMRM. But there’s no reason to expect large changes between the forthcoming release candidate and the full release (at least not at this stage), other than bug fixes. The other members of the @staff team may wish to add to that…(?)


If anything, 3.0_RC3 has more changes than we intended; we’ve been collecting various fixes and minor improvements, which will now go out to master since we want to push the tckgen bug fix. In terms of the “official” MRtrix3 release, it’s primarily about setting up mechanisms for distribution / installation other than git. We don’t have a set of changes that we’re “holding back” for the official release: 3.0_RC3 will contain the most up-to-date finalised code we have for everything at the time at which we push the button.

It shouldn’t be too far away: Indeed 'll probably create the GitHub pull request for it today so that we can start finalising what needs to be done. Unfortunately getting the workshop set up, combined with grant season here in Aus, are dragging things out a little.

Do you expect changes in msmt reconstruction, fod2fixel or tckgen?

MSMT is almost certainly untouched. fod2fixel's -peak and -disp outputs, and FOD-based tractography, will be altered in 3.0_RC3. Another important change that will affect behaviour is the changes to the N4 parameters in dwibiascorrect. I’ll have a better idea of other changes once I create the pull request, since it makes the code changeset easier to navigate in order to write a summary changeset for the public.


That sounds very promising, thank you very much for the detailed information!


Could you shortly explain what the bug in FOD-based tractography is and which changes are to expect in fod2fixel's output?


Hi @Jan_Schreiber,

See this link for the original issue (read into it further to discover the implications):

This pull request fixed the bug at hand, but not some of the further implications with respect to tckgen thresholds:

We now have this subsequent pull request ongoing:

If the final changes to the threshold are agreed upon and implemented, the situation should then end up being fixed in the dev branch, which will quite soon thereafter end up in a release (candidate) on master.