Hi Jamie,
it's impossible to judge surface accuracy from one slice. You need to move forwards and back a couple of slices to see if the surface is curving sharply out of plane.
i attached a few more pics (i have many more, but don't want to clog various inboxes too much).
Can you explain a bit more about what is unique about this volume? I didn't really understand from your email. Is it a single subject MRI? Or the average of some volumes processed using some nonlinear registration?
nothing is all that unique regarding the volume. it is a template that was used with nonlinear registration. i have individual subject warps to and from the template... but in the end, the only unique in terms of FS is that that volume has been smoothed and has a bit of a range in terms of voxel intensities. i termed off N3 in autorecon1, but it tends to white mash the superior aspects of potential surfaces.
i attached a histogram, where WM and GM peaks are pretty apparent (this is the image that went into freesurfer). and selecting the bins of that peak give a decent enough selection of GM. i can sent you the volume if you like / if you think that would facilitate trouble-shooting.
thanks much, jamie.
Bruce
On Mon, 7 Apr 2008, Jamie Hanson wrote:
i think my attachment didn't make it through, the first time round. example pics now attached.
thanks much, jamie.
On Mon, Apr 7, 2008 at 1:10 PM, Jamie Hanson jlhanson5@wisc.edu wrote:
Hi Freesurfers-
I was trying to run freesurfer on a study template and was having major problems with pial surface generation. Orig and Main surfaces seem to be ok, but there are major missing chinks in pial surface. Just to give some background, the template was being using for Diffeomorphic Image Registration, so it is a very unique volume.I know it is likely it is an intensity value issue.I wondered what folks have done previously to correct similar issues?
It appears to be across a massive proportion of the brain, so correcting by hand (by altering intensity values seems a bit labor intensive and perhaps questionable in terms of reliability). I also thought of trying to rerun some of the initial preprocessing (autorecon1 + 2) with some more 'expert settings'. Initially, I was running with no skull strip and no n3 flag (b/c the template has already been skull-stripped and massive preprocessed), but was looking for some additional advice and guidance.
Any thoughts or suggestions would be greatly appreciated.
Best, jamie.
-- Jamie L. Hanson Waisman Laboratory for Brain Imaging & Behavior | Child Emotion
Research Lab
University of Wisconsin - Madison 1500 Highland Avenue Madison, WI 53706 ***Please note my new email: jamielarshanson@gmail.com***