I am working on some data normalization. It seems ‘ok’ at best in some cases (the normalized brains are approximately overlaid though there is still some variability in brain shape in some regions), but not quite right in others. In the not quite right cases, the brain is a bit bigger, and shifted anterior. This seems to happen manly in cases where the top slice still has some signal (we have whole brain coverage, but the top slice might be just on the edge).

Also the offset images are more likely to have a small misregistration with the anatomical T1, but all attempts to get coregister to do a better job have not produced improvements.

The deformation fields (the y_anat.nii file) do not seem correct to me. I have pasted the first two indices of the file as the pics below, and it is a gradient rather than what I would expect a warp field. I thought this may be image inhomogeneity issues, so I tried using homogeneity corrected T1s (produced by freesurer prior to segmenting), same thing.

I wold appreciate an advice on optimizing the normalizations, and what might be going on in these warp fields.