Hi Andrew

the brain.mgz is very aggressively filtered explicitly for cortex and wipes away some subcortical boundaries (e.g. lateral thalamus).

cheers Bruce On Wed, 8 Feb 2017, Andrew Davis wrote:

Hello FreeSurfer Developers,

I have been using FreeSurfer for segmenting T1-weighted images, and I have a question related to using the two outputs norm.mgz and brain.mgz.

It is my understanding that, by default, FreeSurfer uses the norm.mgz volume to do partial volume weighting when outputting segmentation volumes to aseg.stats based on aseg.mgz [^1]. Further, I understand that brain.mgz is output from a second normalization step in the recon-all processing stream, having taken norm.mgz as an input [^2].  The wiki says "Intensity normalization works better when the skull has been removed. Creates a new brain.mgz volume".

My question is: if brain.mgz is meant to be an improved version of norm.mgz, why not use brain.mgz for partial volume weighting?  What advantages does norm.mgz provide?

Thanks much, Andrew Davis

[^1]: https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllDevTable [^2]: https://surfer.nmr.mgh.harvard.edu/fswiki/recon-all

​​​