Thanks, Anja.
We will consider this in future versions, but for the time being, you can use the ugly hack described below.
Best wishes,
/Eugenio

UGLY HACK

  1. Make two copies of recon-all-clinical.sh
  2. In the first copy, remove all the content after the line that reads “# SynthSR”. This will effectively stop the processing after SynthSeg.
  3. In the second copy, remove these lines (which represent the processing made by the first copy):

# Initial mri_convert
set cmd="mri_convert $INPUT_SCAN $SUBJECTS_DIR/$SNAME/mri/native.mgz -odt float"
$cmd |& tee -a $LogFile
if ($status) then
  echo "Error in mri_convert" |& tee -a $LogFile
  exit 1
endif
echo " " |& tee -a $LogFile

cd $SUBJECTS_DIR/$SNAME/mri

# SynthSeg
set cmd="mri_synthseg --i ./native.mgz --o ./synthseg.mgz --parc --threads $THREADS --robust --vol ../stats/synthseg.vol.csv --cpu --qc ../stats/synthseg.qc.csv"
if ("$CT_FLAG" == "true") then
  # If CT, pass --ct
  set cmd = "$cmd --ct"
endif
$cmd |& tee -a $LogFile
if ($status) then
  echo "Error in SynthSeg" |& tee -a $LogFile
  exit 1
endif
echo " " |& tee -a $LogFile

  1. Run the first script. This should be pretty fast.
  2. Manually edit mri/synthseg.mgz
  3. Run the second script. 

This should work because it “intercepts” the script at the first segmentation. 



--
Juan Eugenio Iglesias
http://www.jeiglesias.com
From: Betz, Anja Katharina <Anja.Betz@med.uni-muenchen.de>
Date: Wednesday, August 5, 2026 at 4:56 AM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Re: recon-all-clinical pediatric ventricle segmentation

        External Email - Use Caution        
Dear Eugenio,

you are right, the surfaces should be fine. “Downstream” may have been the wrong wording: There are other outputs of freesurfer (aparc+aseg.mgz, wm.ngz, wm.seg.mgz, wmparc.mgz,…) that are later used in the processing of other modalities (e.g., ASL, DTI/tractography) and often include the ventricles as an explicit label or a standardized value. 

Manually editing only the aseg would mean that all other outputs still contain the undersegmented ventricles, and editing them all separately would create inconsistencies. My question was aimed at a possible way to edit the ventricles once and re-process the other outputs to reflect the changes. 

Best wishes!
Anja 

From: Iglesias Gonzalez, Juan E. <JIGLESIASGONZALEZ@mgh.harvard.edu>
Date: Monday, August 3, 2026 at 20:35
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Re: recon-all-clinical pediatric ventricle segmentation

Dear Anja,
I would argue that, if you manually edit the region, there’s no need to rerun anything downstream since the surfaces won’t be affected, no? All you’d need to do is recomputing the ventricular volume.
Independently: I would quickly run the images (T1 and/of FLAIR) through SynthSeg and see if the ventricles look better. If they do, you can simply take the ventricular masks & volumes from there, right?
Best wishes,
/Eugenio

--
Juan Eugenio Iglesias
http://www.jeiglesias.com
From: Betz, Anja Katharina <Anja.Betz@med.uni-muenchen.de>
Date: Monday, August 3, 2026 at 11:40 AM
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] recon-all-clinical pediatric ventricle segmentation

        External Email - Use Caution        
Good afternoon,

we have acquired a number of pediatric scans (ages 6-18y) on a 3T research MRI scanner. Most likely due to motion artifacts, the regular recon-all pipeline produces (amongst other problems) severe undersegmentation of the gray matter. Instead, we have started using the recon-all-clinical pipeline, which on average seems to produce more accurate results. Our FS version is 8.0.0.

Unfortunately, a reoccurring problem is that the recon-all-clinical workflow often undersegments the ventricles and classifies them as white matter. Two images from two different subjects are attached to show the issue. 

Thank you and best wishes!
Anja