I think it is because n3 sometimes creates a few voxels that are very bright (or can't handle a very bright voxel that is there), then the conform step ends up compressing a lot of voxels into the dark range to handle the few bright voxels. I put some stuff in recon-all to try to reduce the effect.
doug
On 08/08/2013 11:46 AM, Bruce Fischl wrote:
maybe try scaling up the orig.mgz then? I never really understood why the nu darkens it sometime. Alternatively you could scale up the nu, but you will have already lost dynamic range
On Thu, 8 Aug 2013, aqureshi@nmr.mgh.harvard.edu wrote:
The 001.mgz actually looks good in this case (in terms of intensity) so I'm not sure if it makes sense to scale it up further. The problem is with the nu.mgz being too dark. I'll try it out though if you think it will also work in this case.
- Abid
you could try using --scale <scale factor> to scale the intensities of the 001.mgz up
Hi Dr. Fischl,
I had a couple subjects that failed recon-all. Although the error log said recon-all failed at mri_watershed it looks like the root problem is a couple steps earlier, because the nu.mgz looks too dark. This was discussed in a previous thread from 2008.
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg07293.html.
You wrote, "p.s. are the 001.mgz and 002.mgz dark? The orig.mgz? Sometimes nu_correct can create very dark images, and there are some switches you can loo through to correct them. Alternatively if it is the 001.mgz, mri_convert has a scale option you can use."
On my subject 001.mgz looks good (there was no 002.mgz file). It is only nu.mgz that looks too dark. It looks like the solution you gave then was for a problematic 001.mgz file, but from your response it looks as if the problem was nu.mgz the solution would be different. What are the switches you can loop through to correct the nu.mgz.
Thanks, Abid
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer