Dear Ajay,

Please take a look at this paper (which is based on SynthSeg, but SuperSynth is essentially the same method with more outputs and better training data): https://www.sciencedirect.com/science/article/pii/S1361841525002099

In short:

You can run SuperSynth on the timepoints independently, and it does a pretty good job already.
You can get segmentations that are longitudinally a bit cleaner by running USLR on top (see the aforementioned paper).
SuperSynth does not currently have a cortical stream.

Best wishes,

/Eugenio

Juan Eugenio Iglesias

http://www.jeiglesias.com

From: Ajay Sharma <ashar126@jh.edu>
Date: Thursday, July 9, 2026 at 2:23 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] SuperSynth Method and Longitudnal Analysis

External Email - Use Caution

Dear Team,

We are currently using the new SuperSynth method to process and segment 3T T1-weighted MRI volumes acquired at three longitudinal time points (Baseline, Visit 1, and Visit 2).

We would like to know whether it is possible to perform a longitudinal analysis using the outputs generated by SuperSynth. If so, could you please describe the recommended workflow or commands to follow?

Specifically, we are interested in the following:

Can SuperSynth outputs be used for longitudinal processing across multiple time points?
Is there a recommended pipeline or set of commands for longitudinal analysis?
Can we generate cortical thickness measurements and cortical thickness maps to visualize brain changes over time?
Are there any additional recommendations or best practices for performing longitudinal structural analyses using SuperSynth, SynthSeg or recon-all-clinical outputs?

Any suggestions, documentation, or example of workflows would be greatly appreciated.

Thank you for your time and assistance.

Best,

Ajay