Histology atlas has inconsistent origin

[mguaypaq]

Following #6, #7, #8, the histology atlas now has a consistent orientation and pixdim, but 2 of the 7 images have a different origin. As a result, opening the files in fsleyes gives a warning:

Images have different orientations/fields of view - alignment/orientation labels may be incorrect!

Specifically, the affine matrix these 5 files is:

PAM50_200um_AVF.nii.gz
PAM50_200um_Eccentricity.nii.gz
PAM50_200um_EquivDiameter14.nii.gz
PAM50_200um_EquivDiameter48.nii.gz
PAM50_200um_MVF.nii.gz

array([[  -0.19907407,    0.        ,    0.        ,   10.65227222],
       [   0.        ,    0.19879518,    0.        ,  -53.37294388],
       [   0.        ,    0.        ,    0.5       , -184.49050903],
       [   0.        ,    0.        ,    0.        ,    1.        ]])

And the affine matrix for these 2 files is:

PAM50_200um_EquivDiameter.nii.gz
PAM50_200um_Naxons.nii.gz

array([[  -0.19907407,    0.        ,    0.        ,   10.70833302],
       [   0.        ,    0.19879518,    0.        ,  -53.42833328],
       [   0.        ,    0.        ,    0.5       , -184.34002686],
       [   0.        ,    0.        ,    0.        ,    1.        ]])

In terms of voxels, the shift between the two origins of these coordinate systems is approx 0.3 voxels:

>>> (10.65227222 - 10.70833302) / -0.19907407  # x dimension
0.28160774529801463
>>> (-53.37294388 - -53.42833328) / 0.19879518  # y dimension
0.2786254676798286
>>> (-184.49050903 - -184.34002686) / 0.5  # z dimension
-0.30096434000000727

Since I wrote this old comment, I now know how to use nibabel to do pretty much any resampling we want, as long as I know:

• the desired resolution in each dimension (so, it could be exactly 0.2x0.2x0.5mm as previously discussed),
• the origin we want to use (which one?),
• the data array size (currently it's 108x83x205 voxels, but I could also round up to fully cover the current field of view).

Also, I should probably resample starting from the earlier, high-resolution images, rather than the current already-resampled images.

After doing this, what would be the steps to update everything downstream (for example, in SCT)?