Re-running fMRIprep with AROMA

Hi neurostars experts – we are ~3 weeks into running fMRIprep on our data and I have just realised that Aroma–something we wanted to run-- is not run by default (my bad).

It’s a large dataset, and after running some subjects individually, we expect the whole thing to take about 30 days (so we’re kinda close!).

Now, is the best thing for me to do to let this preprocessing finish and then run fMRIprep again but with the Aroma flag? Or should I kill what we’re doing at the moment and start again?

As long as you still have the working directories, it should make relatively little difference between trying to finish first and then re-running with AROMA. If you stop and restart, you’ll get a little boost in that downstream nodes that depend on whether or not you use AROMA may not need to be rerun.

If you don’t have the working directories, you’re going to need to re-run all components (except FreeSurfer), and therefore stopping ASAP and starting over will be your best bet, as any work you do before restarting will be lost.

In either case, I’d recommend stopping now. If you have more specific details you think might be relevant, we can see if there’s anything that would change that, but I doubt it.

1 Like

This sounds surprisingly slow.

  • How many participants were in the study?
  • What are the hardware specs (numbers of virtual cores/CPUs, amount of memory/RAM) of the system you are using?
  • Are you using Docker? If so what resource restrictions are set in settings?
1 Like

Thanks effigies! We will start re-running it now :slight_smile:


Hi Chris – thanks for replying =)

The study N = 79.

We have 8 CPUs, and 32GB of RAM, and have no restrictions on the Docker as far as I can tell.

When I run the command “sudo docker stats”, it says that the Limit is “15.51GB” but the “Mem Usage” value is often exceeding it. While I’ve written this post, it has reachers as high as 21GB. So I suspect that it actually has 32GB and perhaps after our RAM got upgraded recently, the limit hasn’t updated or something?

Do these computational resources suggest it shouldn’t be taking that long?


I see. The estimates check out for this hardware.