In computed tomography, model-based iterative reconstruction relies on an explicit image representation that shapes both the forward model and the ability to recover quantitative features. The choice of basis is therefore not a minor implementation detail.
A new arXiv preprint (2610.24363) argues that for signals that are isotropically bandlimited, the hexagonal lattice offers the most efficient sampling. The authors propose hexagonal basis projectors as a way to bring that efficiency into iterative reconstruction.
The abstract does not provide full results or comparisons, so the practical gains remain to be seen. As a single-source preprint, the claims have not yet been peer-reviewed or independently replicated.