MRIcroGL always orients images to the nearest orthogonal orientation to world space. Note that this ensures voxels are rectangular which is useful for drawing regions and reducing aliasing artifacts. The reported coordinates in mm always use the affine rotation matrix, so they handle the fact that your image may be sliced obliquely relative to world space. I would strongly discourage exporting olbique images resliced to the grid, as this will bake in the aliasing artifacts. This is typically only an issue with raw imaging data: once you normalize an image to standard space, the standard space template will be perfectly aligned with world space (requiring a single, lossy interpolation step).

Our upcoming NiiVue allows you to show data in either voxel or world space. You can drag and drop your data to this live demo:

  https://niivue.github.io/niivue/features...

Note that in the initial example the voxels are rectangular in world space, but the mesh spheres are ellipsoids, while selecting world space makes the voxels rhomboidal but the spheres are round.