Orientation also for non-istropic FOVs and voxel sizes

[ckolbPTB]

Problems with voxel size, shape, size, dimension,... have been mentioned in many issues ( #906, #791, #812, #632,...). A possible approach to test the handling of orientation and geometric information is:

• review exisiting test scripts
• write clear documentation of what get_dimensions(), get_spacing(),... should return
• test self-consistency of functions:
  • create unistropic data sets for MR and PET (number of voxels and voxel dimensions different along all three axis)
  • place offcentre sphere into the image and calculate centre of mass
  • transform image to different formats/orientations and return centre of mass in LPS at each stage
• test consistency with scanner:
  • repeat tests with scanner data which will have the same number of voxels and voxel dimensions along x and y

[DANAJK]

Order of dimensions: Should objects report the order of dimensions of their outputs? This might come as strings such as {'row', 'column', 'slab', 'echo', 'dynamic'}. This has the advantage of not requiring documentation separate from the code, forces users to think about the data they are handling and allows for future expansion. We would need a list of reserved names.