If you run samseg with --mrf, then it will do Markov random field post processing, which will clean up some of that stuff (same post-processing operation that is done when recon-all finishes its segmentation). However, while it made the labels look better, I found that it actually hurt the ability to detect the difference between ADs and controls in hippo volume. doug
On 3/18/2021 6:18 AM, Koen Van Leemput wrote:
External Email - Use Caution
Hi Donatas,
I do think this is really a matter of postprocessing - samseg currently doesn't do *any* of that, although it would probably be beneficial to include some spatial constraints to clean up segmentation errors like this.
Sorry I don't have better news at the moment,
Koen
On Tue, Mar 16, 2021 at 3:08 PM Donatas Sederevicius <donatas.sederevicius@psykologi.uio.no mailto:donatas.sederevicius@psykologi.uio.no> wrote:
External Email - Use Caution I’m currently inspecting segmentation results of putamen from ASEG and SAMSEG, and SAMSEG tends to include claustrum whereas ASEG is more robust to this, see an attached example. Do you think it could partly be due to the fact that volume-based atlas of ASEG is somehow more robust than SAMSEG’s mesh-based atlas given claustrum’s close proximity to putamen or this “minor” mechanical clean up really does its job pretty well? If so, maybe it could be applied to the output of SAMSEG? I’m trying to argue why possibly this could be the reason as both atlases do not contain claustrum. Thanks a lot! -Donatas _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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 <https://secure-web.cisco.com/1qLGmHR1V1Hvprkk4fWfFj2Z0mtiK-DryotGQyo7zSrTIWmmaa-wQarzrPlcG5dvj9pYPtl7QcsKzJ_7Af0_mD6UNIOxia3VVtoXXelC7AbtzqUuhdoZpOp4kcZ4-kCXzd9b2BBhJcRi5cDzbIhjuhPeYN6mLj8iTzMrnLQVfHlQaWa_C6UP4vzl2_Z9HTxQErBYxDJh8VjALBrLCkYf1p8qF0Nnbcy35kk_Dx5GABea2K8q6Tx5e6N160b1WNekm8od6hUxQNTDmOiknXjisYw/https%3A%2F%2Fmail.nmr.mgh.harvard.edu%2Fmailman%2Flistinfo%2Ffreesurfer>
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
