Hi Matt
Yes, about 15% of the population has a folding variant that our old code struggled with. JOSA resolves many/most of those and also significantly outperforms other spherical registration in terms of both folding alignment and functional alignment. This is the bio archives link:
https://arxiv.org/abs/2311.08544
and the Med. Image Analysis published link:
https://pmc.ncbi.nlm.nih.gov/articles/PMC12867183/
here is the relevant figure (4):
So in my opinion better to fix this using the high-resolution registration up front rather than let it be blurred out downstream
Cheers
Bruce
From: Glasser, Matthew <glasserm@wustl.edu>
Sent: Tuesday, August 4, 2026 9:08 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>; Li, Jian,PhD <JLI112@mgh.harvard.edu>
Subject: [Freesurfer] Re: Misidentification of the central sulcus in fsaverage registration
External Email - Use Caution
Hi Donna,
We have a few HCP individuals with this behavior as well. I have one of these cases in a talk I give to neuroradiology trainees. Most of them get it wrong too—it isn’t the easiest problem to solve with folding alone, though it would be interesting if folding registrations trained with multi-modal features like JOSA are better equipped to handle this.
As I see in your ppt. we don’t really worry about this in the HCP Pipelines context because it always gets corrected when we run multi-modal registration.
Matt.
From: Dierker, Donna <donna@wustl.edu>
Date: Tuesday, August 4, 2026 at 3:12 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>;
Li, Jian,PhD <JLI112@mgh.harvard.edu>
Subject: [Freesurfer] Re: Misidentification of the central sulcus in fsaverage registration
Hi Madison, Bruce, and Andrew,
I also have been looking at this issue, since most of the labs I work for use Freesurfer parcels as features in machine learning models.
I'm adding to these slides as I learn more about the issue:
https://wustl.box.com/s/364hrybanrb9zdpcaodk7g8p96lk2nwt
We greatly appreciate your help.
Donna
From: Fischl, Bruce R.,PHD <BFISCHL@mgh.harvard.edu>
Sent: Tuesday, August 4, 2026 10:47 AM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>;
Li, Jian,PhD <JLI112@mgh.harvard.edu>
Subject: [Freesurfer] Re: Misidentification of the central sulcus in fsaverage registration
Hi Mackenzie
This is almost certainly an error in the spherical registration. You can try Andrew Li’s JOSA algorithm, which is likely to fix this. Andrew: can you provide directions on how to run JOSA in the current recon-all?
Cheers
Bruce
From: Mitchell, Mackenzie <macmitch@illinois.edu>
Sent: Tuesday, August 4, 2026 8:35 AM
To: freesurfer@nmr.mgh.harvard.edu
Subject: [Freesurfer] Misidentification of the central sulcus in fsaverage registration
External Email - Use Caution
Hello,
We are seeing that in some cases (13/167 subjects) FreeSurfer v7.3.2 incorrectly labeled the central sulcus in at least one hemisphere (see example below). In all cases, FreeSurfer is
labeling the postcentral sulcus as the central sulcus. In one case, this is happening twice for the same subject who participated in two independent projects (i.e., we find this issue with different T1s and Freesurfer runs). See example below.
We have checked the segmentations and alignments for all subjects and they all look correct. We think this may be happening at the stage of normalization to fsaverage.
Most critically for us, it looks like this error is displacing the mapping of our functional data to the surface, as the motor network functional connectivity maps are shifted forward,
as would be expected from the mislabeling. See example below.
In an attempt to resolve this issue, we have run FreeSurfer v8.2.0 on these 13 subjects. This resulted in the central sulcus being correctly identified in only 4/13 subjects.
Searching through past posts to this mailing list, we found this thread (https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg70250.html)
which suggested to rotate the spherical registration (e.g., z+15 degrees). We found that with various rotation parameters we could improve the cortical parcellation in a handful of subjects. However, we do not currently use the cortical parcellations for our
work and our primary concern is being able to accurately map functional data to the surface. We are not sure if spherical rotation is the right point/method to intervene.
All insight on how to intervene to improve fsaverage normalization would be greatly appreciated.
Thank you,
Mackenzie
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.