Subdivision-based deformable models for extracting volumetric structures from biomedical imagesMS2

An important challenge in biomedical imaging is the characterization of volumetric structures. In a clinical context, the delineation of organs such as lungs and kidneys allows for better 3D visualization and hence facilitates preoperative steps. In a biological context, microscopic images often contain hundreds of cells for which an automatized or semi-automatized cell segmentation is necessary because the manual delineation of each cell would otherwise be overly time consuming. 3D deformable models are powerful tools for the extraction of volumetric structures from biomedical images. They consist in flexible surfaces that are deformed from an initial user-provided configuration toward the boundary of the object to be delineated. The deformation can be driven manually, by interactively modifying the control points of the model, or automatically by minimizing suitable energy functionals. In this talk we describe the surface of a 3D deformable model by the limit surface obtained by applying a suitably defined subdivision scheme to an initial coarse triangular mesh. The benefits provided by the considered subdivision scheme are related to i) its efficiency to characterize 3D biomedical structures with sphere-like topology; ii) its applicability to an initial mesh with very few vertices; iii) its discrete nature that leads to an easy implementation. Finally, it allows for a friendly user interaction whenever some manual editing of the model is desired. Application examples of extraction of volumetric structures from real 3D biomedical images are illustrated.

This presentation is part of Minisymposium “MS2 - Interpolation and Approximation Methods in Imaging (4 parts)
organized by: Alessandra De Rossi (University of Torino) , Costanza Conti (University of Firenze) , Francesco Dell'Accio (University of Calabria) .

Lucia Romani (University of Milano-Bicocca)
Anaïs Badoual (EPFL, Lausanne)
Daniel Schmitter (EPFL, Lausanne)
Michael Unser (EPFL, Lausanne)
computer graphics, image segmentation