Weighting, as implemented, is designed to minimize variance of intra-brain voxels over extra-brain noise. However during the 3dVolreg process motion vectors are created based on signal variance. A given voxel's variance is weighted based upon a blurred mask. I understand from above that this mask normally consists of all voxel's whose signal is greater than 97.5% of the max of the base image. The -weight flag allows us to specify a different mask, but that mask appears to undergo the same thresholding and blurring.
Is there a way to specify the distribution of the weighting function such that we can incorporate more of the variance of the outer brain voxels and less weighting for center brain voxels (r-squared dependence for example) in 3dvolreg?
rob risinger