@effigies The workflow still performs as expected with your changes.

Great! Thanks for the quick feedback.

Unrelated to this PR (except insofar as I want to get rid of the SimpleITK dependency), could I get some inputs and results to center_volume and recodeLabelMap? These should be pretty easy to rewrite with nibabel, as long as I can check my work.

@effigies No problem! Here is an example with inputs and outputs that uses both: https://www.dropbox.com/s/ovmwprr71lcbi87/recodingexample.tar.gz?dl=0
Let me know if you need anything else.

Do you have some inputs that have an affine matrix that doesn't look like this?

1 0 0 a
0 1 0 b
0 0 1 c
0 0 0 1

I want to understand center_volume. Right now it looks like the algorithm just sets [a, b, c] = [±dim1/2, ±dim2/2, ±dim3/2], but that may simply be a consequence of there being no rotation or scaling.

Incidentally, this seems odd to me, and I want to verify that your goal really is just to adjust the affine matrix, and not to translate the contents of the data array such that you have a nice affine matrix. The consequence of center_volume seems to be to put it out of alignment with other files from the same subject. Admittedly, I haven't looked closely at where the resulting files are used, so this could be a dumb question.