confusius.registration.resample_like accepts both affine (numpy.ndarray) and B-spline (xarray.DataArray) transforms, but the wrapper Atlas.resample_like only accepts an affine matrix (src/confusius/atlas/atlas.py:513-520, typed npt.NDArray[np.float64]).
This blocks users who registered an atlas with register_volume(transform="bspline") from resampling the atlas with the resulting transform. The docstring explicitly states that Atlas.resample_like is meant to mirror confusius.registration.resample_like, so the asymmetry is unintentional.
Meshes returned by get_mesh should follow the same deformation as the volume to stay useful as overlays — meaning the affine assumption baked into the mesh-to-physical update (src/confusius/atlas/atlas.py:595) needs to be lifted in the process.
confusius.registration.resample_likeaccepts both affine (numpy.ndarray) and B-spline (xarray.DataArray) transforms, but the wrapperAtlas.resample_likeonly accepts an affine matrix (src/confusius/atlas/atlas.py:513-520, typednpt.NDArray[np.float64]).This blocks users who registered an atlas with
register_volume(transform="bspline")from resampling the atlas with the resulting transform. The docstring explicitly states thatAtlas.resample_likeis meant to mirrorconfusius.registration.resample_like, so the asymmetry is unintentional.Meshes returned by
get_meshshould follow the same deformation as the volume to stay useful as overlays — meaning the affine assumption baked into the mesh-to-physical update (src/confusius/atlas/atlas.py:595) needs to be lifted in the process.