fMRIPrep ERROR: Talairach failed!

Summary of what happened:

I am using fmriprep 25.1.3 to preprocess my dataset. For some of my subjects, I get the following error:

ERROR: Talairach failed!

Command used (and if a helper script was used, a link to the helper script or the command generated):

I used fmriprep with the default options.

Version:

25.1.3

Environment (Docker, Singularity / Apptainer, custom installation):

Linux environment module / Singularity

Data formatted according to a validatable standard? Please provide the output of the validator:

Dataset is bids-conform, no errors, only warnings concerning missing sidecar-keys

Relevant log outputs (up to 20 lines):

	#@# Talairach Failure Detection Thu Sep 18 14:20:26 CEST 2025
	/zi/home/johannes.wiesner/work/slurm/project_sosense/tmp/630476/sourcedata/freesurfer/sub-SOSEIQ4FW/mri

	 talairach_afd -T 0.005 -xfm transforms/talairach.xfm 

	ERROR: talairach_afd: Talairach Transform: transforms/talairach.xfm ***FAILED*** (p=0.0279, pval=0.0034 < threshold=0.0050)

	Manual Talairach alignment may be necessary, or
	include the -notal-check flag to skip this test,
	making sure the -notal-check flag follows -all
	or -autorecon1 in the command string.
	See:

	http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/Talairach


	ERROR: Talairach failed!

	Linux zislhcn0112.zi.local 6.8.0-63-generic #66-Ubuntu SMP PREEMPT_DYNAMIC Fri Jun 13 20:25:30 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux

	recon-all -s sub-SOSEIQ4FW exited with ERRORS at Thu Sep 18 14:20:27 CEST 2025

	For more details, see the log file /zi/home/johannes.wiesner/work/slurm/project_sosense/tmp/630476/sourcedata/freesurfer/sub-SOSEIQ4FW/scripts/recon-all.log
	To report a problem, see http://surfer.nmr.mgh.harvard.edu/fswiki/BugReporting

Screenshots / relevant information:

I checked my T1w images (they are not non-uniformity corrected by the scanner, but this shouldn’t be a problem?) and they look good:


Hi @JohannesWiesner,

This usually happens in images with poorer quality. Have you considered trying running freesurfer outside of fmriprep?

Best,
Steven

Hm that’s strange. When I visually inspect the other images (where this error does not occur), they do not look any better or worse than those where the error occurs.

What benefit would a manual freesurfer run have?

It might finish, I guess, worth a shot.

How do I imitate what fmriprep (with default settings) is doing? Like such?

recon-all -subjid sub-01 -i /path/to/bids/sub-01/anat/sub-01_T1w.nii.gz -all

Yah, unless you also have T2 images.

1 Like

I have a total of four subjects with this error. For two of them, manual reconstruction using recon-all seemed to have “worked” (i.e., I still see the error in the Freesurfer log, but the reconstruction seemed to have continued. However, I do not get the exact number of files that I get for the other subjects that were preprocessed using fmriprep). For two of the subjects, however, the manual reconfiguration continues to fail. Do you have any other ideas about what I could do? This error seems to be covered quite extensively elsewhere.

See, for example (but there are other threads as well):

I could run recon-all without errors on these subjects with:

recon-all -subjid $SUBJECT -i $INPUT -all -parallel -openmp 16 -notal-check -skullstrip -no-wsgcaatlas

However, the output was really bad (missing surfaces for a lot of brain regions and/or artifacts like surfaces out of the cortex and or pials that belong to the wrong hemisphere). I think that’s all due to the non-uniformity in my T1 images (see above, the occipital areas are brighter than the rest of the brain).

Still strange, because I don’t find that the input images have drastic artifacts. Also found this article:

However, FS 7.1 underperformed due to intensity normalisation problems leading to the exclusion of a substantial part of the occipital cortex from the brain mask in 6 subjects (4 HCs and 2 T1D patients)

I currently run fmriprep on these subjects without the recon-all workflow

I ran fmriprep sucessfully on all my subjects with fs-no-reconall. I also checked all the brain masks and they look much better than the ones created by recon-all (i.e. even for the subjects for which I didn’t obtain any errors the brain masks from recon-all sometimes missed out smaller parts of the occipital lobe).