Hi AFNI-Experts!
I wonder, if the following masking procedure would be reasonable and would be ok for any further statistical analysis:
The EPI-mask is based on the deconvolution error-output. Since the standard error is apparently high at the edges of the brain probably due to partial volume effects, it would be desirable in my point of view to exclude these voxels of any smoothing procedure across subjects. I calculated the MAD (with 3dTstat) of the error 3D+time BRIK and zeroed all voxels out higher than 2.5, which looked quite good in my datasets.
I wonder, if anyone sees any problems in that procedure?
Thanks,
Lukas
