Differences between machines, same version fMRIprep

1

Dear all,

We ran one subject through fmriprep version 21.1.0 (docker version) on two different machines. This provides different output, it’s almost the same (e.g., correlation between voxel time courses of 0.95), but not fully. Do you have any ideas what causes this difference?

Thanks!

Best
Hans

2

Hello,

Some multi-threaded processes will produce slightly different outcomes between runs, even if the RNG seed is constant. For fMRIPrep to be fully (or at least more) reproducible, one would need to set the random seed to be constant and be single threaded, which will be very slow.

Best,
Steven

3

Hi Steven,

That makes sense. Thanks for your swift response.

Best
Hans