Preprocessing with different versions of MRtrix


I have preprocessed a big batch of dwi data using mrtrix3 (running dwidenoise, mrdegibbs, dwipreproc, and dwibiascorrect), and have recently had to change machines prior to running the next batch of participants through. I have a new install of mrtrix3.0.2 on this new machine, and can see that a couple of commands have been updated (e.g. dwipreproc is now dwifslpreproc). I have checked the source code for any big discrepancies, but just wanted to check with the developers whether it should be safe to continue preprocessing (and only preprocessing- tractography is done on another machine) with the new version, or if I should attempt to build version 3 on the new machine from source? I guess I am wary of introducing artificial differences into the data.

Thank you very much!


Hi Caitlin,

Unfortunately it’s hard to say for certain whether any of the many changes that will have been made to the code since your previous version will affect your results. I would recommend you take a look through the change logs for each version increment and see whether the commands you’re relying on are mentioned – you’ll find them all in the announcements category. I expect the renaming of dwipreproc to dwifslpreproc won’t have made much difference, but the specific version of FSL that the script invokes will. Other than that, I suspect there may have been changes to dwidenoise that could potentially affect the preprocessing as well.

Probably not what you wanted to hear, I’m afraid…
All the best,

Not exactly what was hoped for @jdtournier, but as I had expected, so thank you for confirming!

Actually, the referral to the announcements page highlighted somewhat of a larger problem…it seems as though I have been using a dev release that I should not have been (version 3_RC4). I am now thinking the best idea is just to re-run everything using the most up to date software? Thank you again for your help!

Sorry, again, it’s really hard to provide any guarantee here. I don’t anticipate there would be major differences in the preprocessing between these two versions, but there may be some small differences, and I’ve no idea whether these would be sufficient to bias your results. All I can reasonably recommend is to run the preprocessing again with the latest version.

Hopefully you’ve scripted everything and it’s just a matter of leaving the computer to it…? :grimacing: