Hi Graham,

If the voxels that were added are of a value 255 and in a location where gray matter was intended to be extended I don't think the edits were done in the right way. It sounds like voxels have been added to that volume when really you want them to be cloned into to it. This puts them in directly from some other volume (i.e., the T1) and keeps the intensities etc consistent. It's basically just using a different feature of the "Edit Voxels" tool. You can see a description here: https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/SkullStripFix (at the bottom under manual editting the Cloning tool is described!)

Jenni

On Wed, 13 Sep 2006, Graham Wideman wrote:

Jenni:

Thanks for your reply and explanation of the brainmask logic and rationale.

As for this:

...

...

Question 1: Is there a rationale for 255 values? (in the brainmask volume)

I'm not sure where these are coming from in your volume - I assume you havent done any editting on this one.

These subjects' data may well have been edited (not by me, hence my not knowing exactly what's what here). And I suppose it's the brainmask.mgz volume that would have been edited, and the act of editing it sets voxels to 255 perhaps? (The 255's are in the appropriate locations to add additional grey matter to the edges.)

That leads me to believe that the answer to this:

...

Question 0: Is the detailed image data in brainmask.mgz ever used, or is it simply a zero/not-zero mask? Ie: is the meaning "masked brain", or "mask using non-zero values to show where brain is".

... is that brainmask.mgz is used by subsequent steps as a mask in combination with some other image volume providing the actual image. In that case the actual intensities of brainmask.mgz are ignored.

And I now see in the script that mri_ca_normalize does indeed use brainmask.mgz with the mask flag, so I guess I'm probably right that the actual intensities are unimportant. (And nu.mgz provides the actual image).

Phew, I think I now have that straight... for this week :-)

Graham