Dear community and developers,
I am experiencing issues during the normalization step, converting NU.mgz to T1.mgz. Gray matter is falsely recognized as white matter and high intensity WM is ignored. The resulting T1.mgz collapses the wrong voxels into the 110 bin.
It seems that what mri_normalize assumes a peak in the WM intensity distribution that is too low. In NU.mgz WM seems centered around 140 not 110. Note that this is not due to RF-field inhomogeneities. It is likely of physiological origin
- Is there a way to explicitly tell mri_normalize to use a different intensity peak?
I realize that correction can be done with control points, however I understand that does not compute a new intensity distribution, but merely adds voxels around CPs. This is undesirable since I have >20 scans to process. Also this will not correct the falsely tagged GM voxels.
I've tried to trick mri_normalize by adjusting NU.mgz intensity by -30 (140-30 = 110). This did not give me the wished-for effect. - Is it possible to bypass preprocessing steps of mri_normalize before collapsing?
- In general it would help if there is some elaboration on the options: -no1d, -nosnr, -gentle, -f vs -fonly, -prune, -g, -monkey - is [-monkey] just shorthand for [-no1d -n 1]?
Thanks, Tom
System: Mac OS X 10.10.5 freesurfer-Darwin-lion-stable-pub-v5.3.0
mri_normalize --version
stable5