Output different when applying topup/eddy separately instead via preproc script

Dear MRtriX Team,
my preproc script outputs are slightly different compared to the results I got when applying topup/eddy separately with Fsl (see picture 1, 2). For both images no additional preprocessing was applied (I know the name of the image is confusing, however no denoising was applied). Is there a reason for the different intensity values? Is the newest version of Fsl implemented in the preproc script?

Thank you for your support!
Best
Max

I’m not sure what would have introduced the difference, but you should certainly be able to see exactly what steps were taken by dwipreproc as it was run. That might help to narrow down where the discrepancy might have been introduced?

Also, to allow us to get to the bottom of it, you’d need to tell us as much as you can about what you did, both on the MRtrix3 side, and on the manual FSL side. There’s a lot of options within dwipreproc… Ideally, it would be best to have the full command-line and coresponding terminal output for all the commands involved.

1 Like

Hi Max,

I think that this is expected, because eddy is not deterministic, so even running the same command in the same data twice, could give you a slightly different results. If you want to be sure to reproduce the same results you have to add the flag --initrand

I hope this helps.

Regards,

Manuel

2 Likes

According to my experimentation with eddy_cuda, with --initrand I still get sligtly different results (although with much lesser variance than without this parameter). There is apparently some indeterministics still left, probably due to the CUDA multi-thread implementation.

Antonin

2 Likes

Okay, I was expecting something like that. Just wanted to be sure my outputs are not wrong.
Thank you for your help!
Best
Max