Challenges in global analysis of high-resolution volumes: complete vasculature estimation and detailed multi-organ labelling
Advances in 3D imaging technologies allow detailed understanding of
the global arrangement of multiple structures/organs,
but computational challenges are also raised. We discuss two
biomedical problems that either require such
detailed analysis of the images or benefit from it.
In particular, we consider extraction of complete vascular tree structure
in micro-CT volumes with near-capillary level resolution.
While discussing standard approaches, we emphasise new ideas using curvature-based
regularisation of the centerlines. We also consider common segmentation problems
in MRI or CT volumes, where similar appearance of tissues makes it difficult to
discriminate organs independently of the context. We advocate simultaneous global
analysis of the data using anatomy-based partially ordered labelling.