Will fMRIPrep 25.x work with FreeSurfer 8.x?

Software Support
,
1

Hello,

I’m wondering if fMRIPrep 25.x will accept precomputed FreeSurfer 8.x subject directories (via --fs-subjects-dir). The container (25.x.x) includes FreeSurfer 7.3.2 — so I suspect that fMRIPrep might not recognize 8.x and could try to rerun recon-all?

I would appreciate any insights or experiences on whether this setup would work.

Thanks!

2

Hi @sreeniva,

As long as the file output structure has not drastically changed between versions it should work fine. If it does throw an error you can try with the --fs-no-resume option.

Best,
Steven

3

Hi Steven,

Thank you so for the response. I am running fMRIPrep using FS 8.1 data, both with and without the -fs-no-resume flag, and have not errored out yet.

The run (run1) without the -fs-no-resume flag invoked a new recon-all using freesurfer 7.3.2 in the container (based on recon-all.log in the folder mentioned via --fs-subjects-dir), while I do not see this for the other run (run2 with the -fs-no-resume flag).

I do not see any modifications in the FS folder for run2. However, the T1 summary for both shows that freesurfer is run by fMRIPrep. So would that mean the FS8.x is not being recognized, but since the structure is similar/same, the processing might go through since all the files are present (?). Thanks again.

4

If you do not see evidence of recon-all being executed in the output logs, then you should be good in assuming fMRIPrep is not rerunning anything.

5

Tried a few weeks ago, i can confirm that it works without --fs-no-resume .

I actually compared running fMRIprep 25.2.3 on freesurfer 8.1.0 vs freesurfer 7.4.1 precomputed output. When ran on freesurfer 8.1.0, it took 1-2 hours longer to complete, there were some additional anat processing which wasn’t required if i ran on freesurfer 7.4.1

for what’s its worth, i also ran fMRIprep on a manually prepared environment with a freesurfer 8.1.0 installation, without any precomputed freesurfer output— it crashed out somewhere during the anat/sMRIprep processing. This wasnt an issue with freesurfer 7.4.1

So i’m not sure if its worthwhile to upgrade to freesurfer 8.x at the moment if you are using it primarily for fMRIprep

6

Hi,
Thank You for sharing Your experience. I am surprised with the result as FS8 is significantly faster than the previous versions. Did You account for total time of FS + fmriprep or just fmriprep analysis?
B.

7

my experience with recon-all on a typical subject from the healthy brain network, without -T2pial:
fs8.1.0 ~3.5 hours
fs7.4.1 ~ 5.5 hours

With -T2pial, the difference in completion times betwen both fs versions narrowed down quite substantially , with fs8.1.0 completing only around 30min earlier than fs7.4.1

so if you add up the recon-all (done outside of fmriprep) and fmriprep completion times for fs7.4.1 and fs8.1.0, i don’t think there would be a significant difference in the completion times

8

Hi gerardyu,

Thank you so much for sharing your experience. This is good to know. Maybe I should give it a shot with 25.2.3 (I used 25.1.3, sorry for not mentioning this information in my first post).

Both my run1 (without -fs-no-resume) and run2 (with -fs-no-resume) finished without any errors. I checked the lastcall.build-stamp.txt for both fsdirs, and for run1 it was fs ver 7.3.2 (indicating a new invocation of fs), and for run2 it was fs ver 8.1.0. (autorecon1,2and 3 were seen in the output log for run1, but not for run2)

Interestingly, run1 took only about 15 minutes longer than run2, despite having to run freesurfer (I ran them almost one after the other). I was expecting a little more time difference. Did you just compare the output logs to find this extra anat processing while using fs8.1, or were there any different output files? Just want to check with my runs as well. Thank you again.

9

unfortunately, no i didn’t look at the logs or the output files in anat/ . I’d imagine the output files will be the same no matter which version of fs was used.