External Email - Use Caution        

Hi Stefano,

For samseg isn't there also an option to disable the VAE shape model? I forgot the details, but if false positives are not an issue maybe that could also help?

Koen

On Friday, January 5, 2024, Stefano Cerri <ste.45.ste@gmail.com> wrote:

        External Email - Use Caution        

Hi,


For SAMSEG:

Are you using the --lesion-mask-pattern flag with the value "1" on the FLAIR scan? You can try to relax the masking by excluding darker than WM mean intensities voxels rather than darker than GM ones, i.e., by adding to your command: --lesion-mask-structure "White"

For WMH-SynthSeg:

It looks like a memory error (the process was killed). On the Wiki: MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://surfer.nmr.mgh.harvard.edu/fswiki/WMH-SynthSeg they recommend 32GB of RAM. Do you have access to a machine with more RAM?

Stefano

[Freesurfer] WMH-SynthSeg vs. SAMSEG for WM lesions

SudEEpti Thu, 04 Jan 2024 23:54:04 -0800

        External Email - Use Caution        

Hello Freesurfer
I am reaching out for some advice on unsupervised WM lesion detection using
SAMSEG and WMH-SynthSeg.

*SAMSEG: *
Our use case prioritizes maximizing true positives over minimizing false
positives. However, even with a low threshold of 0.1, SAMSEG misses some
areas (on FLAIR). Are there any strategies or parameters to force SAMSEG to
be more "aggressive"?

*WMH-SynthSeg:*
Unfortunately, I have not been able to get WMH-SynthSeg to work. The error
message is pasted below. Does this point to a memory issue (I have 24GB
RAM)?














*Arguments seem correct; loading Python packages...Using cpuUsing 1
thread(s)Preparing model and loading weightsWorking on image 1 of 1:
/home/localuser01/southekal/TEMP/101-001_3scans_FLAIR/101-001-1_reg.mgz
 Loading input volume and normalizing to [0,1]     Upscaling to target
resolution     Pushing data through the CNNKilledreal 3m33.897suser
2m48.028ssys 0m32.337s*

Any insights or suggestions would be greatly appreciated! Thank you in
advance.

_______________________________________________
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
Please note that this e-mail is not secure (encrypted).  If you do not wish to 
continue communication over unencrypted e-mail, please notify the sender of 
this message immediately.  Continuing to send or respond to e-mail after 
receiving this message means you understand and accept this risk and wish to 
continue to communicate over unencrypted e-mail.