Hi John, To my knowledge, there is no mechanism of manual editing that can force the pial surface out further (provided that it wasn't a problem of missing pial tissue due to an overly aggressive skull-stripping).
cheers, Mike H. > By the way, here is a matching picture of T1.mgz. Just to show you that > there were really no problems that I could see with any of the > skullstripping - thought FS did fine in that regard. > > -John > > On Tue, Oct 7, 2008 at 1:05 PM, John Sheppard < > [EMAIL PROTECTED]> wrote: > >> Hey Freesurfers, >> >> I have a question regarding segmentation in autorecon2. I have a bunch >> of >> sagittally-acquired MRI data -- Freesurfer has done a decent job after >> we >> manually added control points on the data and resegmented. The problem >> is >> that FS ends up leaving out some lateral areas of cortex. After control >> points, I am satisfied with the grey/white boundary but the pial surface >> could stand improvement. I am attaching an image to this email showing >> a >> segmentation in the coronal plane. Seems like a chunk or lateral >> temporal >> cortex is missing in the left hemi (and a little bit of missing cortex >> in >> right hemi too.) I turned the brightness up a bit so you could see the >> missing cortex in brainmask.mgz -- the lateral stuff as you can tell is >> not >> quite as intense as the more medial voxels, which is probably the root >> of >> the problem. >> >> First of all, I want to get a sense of what is reasonable to expect. >> Does >> this look like a reasonable segmentation or is it reasonable to try and >> improve this? My PI is concerned about the missing lateral GM in any >> case. >> >> Second, I was curious if any experienced users had advice regarding ways >> to >> get this missing cortex included in segmentation - specifically the >> situation in the left temporal lobe. Going off the website, since the >> white/grey surface looks correct, control points should not be used. (I >> don't know whether putting a few *right* outside the surface might help >> expand pial surface? I might try that more if you think it could help..) >> The >> guide would then suggest that pial surface edits must be made manually >> on >> every slice prior to re-running autorecon2-pial. This would take a long >> time for 1 person with 34 brains, but is certainly a possibility. >> Downside >> to this is it becomes somewhat subjective for someone with only a >> rudimentory understanding of neuroanatomy. >> >> What if I could run the autorecon2 workflow step-by-step and try >> adjusting >> normalization parameters? There are two normalizations here, using >> mri_ca_normalize (-canorm) and mri_normalize (-normalization2). >> mri_ca_normalize has a flag -p, that specifies percentage of likely wm >> to >> use as control points (defaulting to 50%). Bumping this up may improve >> intial segmentations since we've had to extensively add control points, >> though I suppose the end result would be about the same (in theory). >> mri_normalize has a number of parameters that might effect this, as well >> as >> various programs called in the segmentation process. >> >> Long story short, does anyone know of any parameter adjustment approach >> to >> this, or is manual editting of pial surface the only solution? >> >> Thanks! >> >> John Sheppard >> >> >> >> > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
