Hi, sorry for the delay. For etiv, the algorithm depends on the linear registration of the brain to the mni305 (ie, it is not based on a segmentation to compute the TIV -- that is what the stiv and it is probably better than the etiv). With v8, we changed algorithms for this computation, and so it is not surprising that there are differences. In my evaluation, the difference was typically about 5%. It looks like you are seeing a much larger difference. Have you checked the registrations to see if they look ok? 

For the others, what command did you use? We actually have two versions of this software (one matlab-based and one python), so I want to make sure you are not comparing against those two versions. The underlying software for the hippo/amyg subfields did not change from 7 to 8. However, those segmentation make use of the hippo and amyg segmentations in the aseg, which is now performed by synthseg instead of the gca. My guess is that this is the source of the change. Why is the change so big? Not sure, but these segmentations are very difficult as there is not really any contrast at the edge of the subfield ROIs and the ROIs are very small. Because the ROIs are so small, it does not take much of a change to get big volume changes.



On 8/10/2025 11:47 PM, Hana McMahon wrote:

        External Email - Use Caution        

Hi Douglas,

Thank you for your reply. The regions I am primarily interested in are the segmentations of the Hippocampus and Amygdala, as well as eTIV as a covariate. However, I will also run exploratory analyses on the whole brain, as the study is a clinical group comparison. 

Regarding some numbers, please see attached (and in the .zip):
  1. roi_proportional_changes: for each ROI (whole brain) I ran the formula mean(abs((v8 - v7) / v7) to determine the average absolute proportional difference between the versions, which is where I got the ~ 1-200% differences between versions in different ROIs
  2. etiv_differences: this is the eTIV (estimated total intracranial volume) for each participant in my study directly taken from the output of each version of FreeSurfer
  3. amyg_differences: this is an example of an amygdala subfield (Medial Nucleus LH)  for each participant in my study directly taken from the output of each version of FreeSurfer

If you could please kindly explain a bit more about why there are big differences between v7 and v8 that would be very helpful, as I am currently not sure which version to move forward with, as the findings are completely different. I was under the impression that v8 is meant to be better, particularly for segmentation, however I am worried about publishing and reviewers, particularly if the outcomes cannot be compared to earlier versions and therefore the current literature.

Thank you again for your help and response.

Kindest,
Hana 

 

From: Douglas N. Greve <dgreve@mgh.harvard.edu>
Date: Monday, 11 August 2025 at 8:36 am
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Re: Large discrepancies between FreeSurfer v7.2.0 and v8.0.0 outputs
You don't often get email from dgreve@mgh.harvard.edu. Learn why this is important
Please be (much) more specific. What are the regions that you are finding changes in? Can you provide some numbers? There are big differences between v8 and v7 so it does not totally surprise me.

On 8/7/2025 1:40 AM, Hana McMahon wrote:
        External Email - Use Caution        
Hi,

I’ve processed the same set of T1‐weighted scans with both FreeSurfer v7.2.0 and the newer v8.0.0. When I compare the regional stats (aseg + subfield tables), the absolute proportional differences are surprisingly large (anywhere from ~1 % up to nearly 200 %, depending on the structure). 

I first noticed the issue while running ANCOVAs: with the v8 outputs, estimated total intracranial volume (eTIV) is highly significant in every model, whereas the v7 results (using identical R code and covariates) show a very different pattern.

I am wondering:
  • Are these version‐to‐version shifts expected, given the changes in v8?
  • Have others reported similar magnitude differences, particularly for eTIV and subcortical/hippocampal measures?
  • Are large shifts expected from run-to-run as well?

Any insight would be greatly appreciated, as I am concerned by the magnitude of the differences I am seeing between versions, which are significantly changing my results. 

Thanks for your time. 

Kindest,
Hana

 


Confidential communication - This email and any files transmitted with it are confidential and are intended solely for the addressee. If you are not the intended recipient, please be advised that you have received this email in error and that any use, dissemination, forwarding, printing, or copying of this email and any file attachments is strictly prohibited. If you have received this email in error, please notify me immediately by return email and destroy this email.



_______________________________________________
Freesurfer mailing list -- freesurfer@nmr.mgh.harvard.edu
To unsubscribe send an email to freesurfer-leave@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman3/lists/freesurfer@nmr.mgh.harvard.edu/



_______________________________________________
Freesurfer mailing list -- freesurfer@nmr.mgh.harvard.edu
To unsubscribe send an email to freesurfer-leave@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman3/lists/freesurfer@nmr.mgh.harvard.edu/