Hi Dana, that is an impressively hard image. I think you'll need to do a fair amount of manual editing to improve the segmentation (and remove topological defects) in the temporal lobe.
Bruce On Thu, 15 Oct 2009, Dana W. Moore wrote: > Hi Nick, > > After trying all of the steps below, the surfaces are much improved, but the > pial surface still has problems, particularly in the temporal lobes (see > below). So, maybe this is the best we can do with this image. Thank you for > helping out with this. > > Thanks, > Dana > > [] > > > At 06:58 PM 10/5/2009, Nick Schmansky wrote: >> Dana, >> >> First I would run: >> >> recon-all -s subid -clean >> >> to remove any prior control points and edits. >> >> Then I would open nu.mgz: >> >> tkmedit subjid nu.mgz >> >> and *sparingly* put control points well into white matter areas on a few >> different slices. you only need a half dozen or so. to add control >> points, see: >> >> http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/ControlPoints >> >> except you will add them to nu.mgz, to make sure that T1.mgz gets >> created properly. >> >> Then create a file name xopts.txt containing these lines: >> >> mri_normalize -gentle >> >> Then run: >> >> recon-all -s subjid -autorecon1 -expert <path_to>/xopts.txt >> >> and when done, open brainmask.mgz >> >> tkmedit subjid brainmask.mgz >> >> and check again to get a sense of the low value for white matter and >> high value for gray matter, then run: >> >> recon-all -s subjid -autorecon2 -seg-wlo wlo -seg-ghi ghi >> >> replacing wlo and ghi with those values. >> >> if the surfaces still look bad, you can add a line to the xopts.txt >> file : >> >> mris_make_surfaces -max_gray val -min_gray_at_white_border val >> >> replacing the two 'val's with something you find appropriate, then run: >> >> recon-all -s subjid -autorecon2-pial -autorecon3 \ >> -expert <path_to>/xopts.txt >> >> good luck! >> >> Nick >> >> >> On Mon, 2009-10-05 at 16:56 -0400, Dana W. Moore wrote: >> > Hi Nick, >> > >> > Below is Bruce's response to a problem I'm having with the gray/white >> > segmentation. (Pasted below, you can see the brainmask.mgz volume and >> > surfaces--the entire brain is labeled as white matter). I could use >> > some guidance in carrying out his advice. He says to adjust the >> > intensity normalization using control points and then use expert opts >> > for mri_segment and mris_make_surfaces. >> > >> > When I look at the brain.mgz volume, the white matter tends to have an >> > intensity of 110, as it should, but the gray matter at times goes as >> > high as 115. I was thinking, I could set ghi to 115 and wlo to 110. >> > Then, I am not sure how to actually run the corrections and what >> > language to use to implement the control points and mri_segment >> > adjustments. I will save control points and then run: recon-all >> > -autorecon2-cp -autorecon3 -subjed <subject>. Can I somehow include >> > the mri_segment adjustments into this command, or do I have to run >> > that separately? >> > >> > I am also not sure how options for mris_make_surfaces might help. >> > >> > I appreciate your help with this! >> > >> > Thanks, >> > dana >> > >> > >> > >> > >> > [] >> > >> > At 04:40 PM 10/2/2009, Bruce Fischl wrote: >> > > Hi Dana, >> > > >> > > it looks like the gray/white density estimation failed. Try setting >> > > them with the expert opts for mri_segment and mris_make_surfaces >> > > (Nick can point you in the right direction if you can't figure it >> > > out). Things like max gm at white border and such. >> > > >> > > cheers, >> > > Bruce >> > >> > >> > p.s. it also looks like the intensity normalization went too far due >> > to the low contrast. You'll probably need to add some control points >> > and run it with the -gentle option >> > >> > >> > >> > > On Fri, 2 Oct 2009, Dana W. Moore wrote: >> > > >> > > > Hi everyone, >> > > > >> > > > I am trying to salvage some images that were done with poor >> > > > gray/white boundary contrasts. First, FreeSurfer includes an area >> > > > of dura in the skull strip. I tried adjusting the watershed but >> > > > it made no differences. FreeSurfer is subsequently unable to >> > > > detect the gray/white boundary and labels the dura as cortex: >> > > > >> > > > [] >> > > > >> > > > >> > > > >> > > > Looking at the raw images, the gray/white boundaries are faint but >> > > > visible. Is there anything I can adjust to try to make this work >> > > > with FreeSurfer? >> > > > >> > > > Thanks, >> > > > Dana >> > > > >> > > > >> > > > Dana W. Moore, Ph.D. >> > > > Neuropsychology Fellow >> > > > Cornell Neuropsychology Service >> > > > Weill Medical College of Cornell University >> > > > New York Presbyterian Hospital >> > > > Department of Neurology & Neuroscience >> > > > 428 East 72nd Street, Suite 500 >> > > > New York, NY 10021 >> > > > Phone: 212-746-2823 >> > > > Fax: 212-746-5584 >> > > > Email: dwm2...@med.cornell.edu >> > >> > Dana W. Moore, Ph.D. >> > Neuropsychology Fellow >> > Cornell Neuropsychology Service >> > Weill Medical College of Cornell University >> > New York Presbyterian Hospital >> > Department of Neurology & Neuroscience >> > 428 East 72nd Street, Suite 500 >> > New York, NY 10021 >> > Phone: 212-746-2823 >> > Fax: 212-746-5584 >> > Email: dwm2...@med.cornell.edu >> > >> > > > Dana W. Moore, Ph.D. > Neuropsychology Fellow > Cornell Neuropsychology Service > Weill Medical College of Cornell University > New York Presbyterian Hospital > Department of Neurology & Neuroscience > 428 East 72nd Street, Suite 500 > New York, NY 10021 > Phone: 212-746-2823 > Fax: 212-746-5584 > Email: dwm2...@med.cornell.edu > _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
