Hi Erik

we try very hard to accomodate a wide variety of T1-weighted MRI sequences with extremely variable contrast properties. That piece of code tries to estimate and adapt to the contrast it finds in the image, but things like subjeect motion can mess it up. If you have a fixed acquisition with known contrast properties then there should be no problem with simply telling recon-all what to expect instead of having it discover it on its own

cheers Bruce

On Fri, 14 Feb 2014, Erik Lindberg wrote:

Dear Bruce and Freesufers,

Adding to our previous discussion. You managed to fix a very bad segmentation by setting some expert options (mri_normalize –gentle and setting a fixed level for -min_border_white  -min_gray_at_white_border).

It was a very nice recovery of, what I thought to be, a hopeless image.  As I in my project have around 800 images I was also testing these options on 10 more images that you used to “recover” the bad image.

The test on the 10 other images (that also worked well without expert options) was quite good. The segmentation looked ok.

What I am wondering now is if there is other aspects to consider when setting fixed levels in mris_make_sufaces? As I told you I looked into the recon-all log file and found that when FS is running normally in creates different levels of -min_border_white  -min_gray_at_white_border for each image.

As this is the Normal way for FS to run I suppose that this is the optimal way to run. So what is the disadvantages of setting fixed levels for -min_border_white  -min_gray_at_white_border?

Thanks!

Erik