External Email - Use Caution        

I am not part of Jee Won's team: just generally interested. For my work, I am looking at ventricular volume and so it would be good to not have to manually correct a large number of my scans.
From: Douglas N. Greve <dgreve@mgh.harvard.edu>
Sent: Tuesday, January 4, 2022 1:19 PM
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: [EXTERNAL] Re: [Freesurfer] Ventricle segmentation issue - ventricle mislabeled as white matter
 

CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and are confident the content is safe.

Hi Jee Won and Priyanka, there are several things going on here. First, the original aseg (aseg.presurf.mgz) has these areas labeled as nothing (0); they show up as WM in aseg.mgz and aparc+aseg.mgz. Obviously, it would be good to have them properly labeled, but is this critical to your applications? Eg, are you looking at ventricular volume?

On 12/21/2021 9:19 AM, Kang, Jee Won wrote:

        External Email - Use Caution        

Hello all,


I found a very common error in our data while I was correcting recons. The lateral ventricles (usually the horns) are being misclassified as white matter in aseg.mgz. This happened in around 1/3 of our 108 subjects.


We were able to fix this issue by manually filling in the ventricles in aseg.presurf.mgz and re-running Freesurfer with the command autorecon2-noaseg -autorecon3, but this is obviously quite a labor-intensive fix for a substantial portion of our data. This was also the only way to fix the segmentation - filling in the mislabeled areas as 255 in the wm.mgz did not fix the problem. I'm not sure why Freesurfer is having issues here, as the areas it is mislabeling have very distinct intensity boundaries - the T1 intensity is usually between 20 and 45 in the areas that are being mislabeled. I've attached a screenshot of a subject's aseg.presurf.mgz who had this ventricle segmentation issue, and you can see that a large portion of the right ventricle has just not been segmented properly; the intensity of the area in the aseg.presurf.mgz and wm.mgz volumes is just 0, and it was 41 (R cerebral white matter) in aseg.mgz.


Would you be able to provide some insight into why this might have been happening, and what we could do in the future to prevent such extensive manual editing in the future? I can also say that this data was processed with a T2 image as well, and supplying that flag didn't help with the segmentation issue.


Thank you,

Jee Won


_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer