Hi Ana,
that could be a problem as you don't have many bits of dynamic range if the image intensities are so low. You could try omitting the nu step and see if it resolves the problem.
cheers, Bruce
On Wed, 23 Sep 2009, Ana Catarino wrote:
Hello,
I was just running autorecon1 on a subject and I got the mri_watershed Error: GLOBAL region of the brain empty!
I am aware that there was a bug in previous versions but I'm using v4.3.1, which I think it's not supposed to have that bug any longer. I tried running -skullstrip with the -no-wsgcaatlas flag, which caused the whole brain and part of the head to be striped out instead of just the skull. Next I tried to compare the output images step by step with images at the same stage for another subject with no error. Regarding the 001.mgz, rawavg.mgz and orig.mgz there were no obvious differences between both subjects, except that the images looked a little darker for the subject that got the error later. However, the nu.mgz image was quite different. The subject with error presented a very dark image, with signal intensities around 40, whereas the subject without error presented normal intensities around 110.
The quality of the original images was the same, same sequence, everything. I already ran 20 subjects and this was the first one to show this error.
Any help provided with be very much appreciated!
Thank you,
Ana Catarino