I have encountered the problem that chrisadamsonmcri describes in issue 898. The last time I used mri_binarize was using FS v6, where it behaves like this:

mri_binarize --replace A B --replace B C

Label A would become B in the output.
Label B would become C in the output.

here's an example:

--replace 11 15 # map "4th Ventricle" -> "4th-Ventricle"\
--replace 15 72 # map "5th Ventricle" -> "5th-Ventricle"\

i want to convert a seg where the LUT maps 11 as 4th ventricle and 15 as 5th ventricle. i want to convert these to the FS LUT (using the replacements shown, 15 and 72). But this results in 4th ventricle getting labeled as 5th ventricle! Four years ago i ran a script performing 230 replace functions to map Neuromorphometrics labels to FS, and it worked fine (evidenced by Dice testing and inspection). Today i ran the same script on the same data and get funky results on a half dozen or so structures that follow the example replacement scheme. My current workaround is to use mri_binarize from v6.