Medial Models Incorporating Object Variability for 3D Shape Analysis
IPMI '01 Proceedings of the 17th International Conference on Information Processing in Medical Imaging
Shape versus Size: Improved Understanding of the Morphology of Brain Structures
MICCAI '01 Proceedings of the 4th International Conference on Medical Image Computing and Computer-Assisted Intervention
Symmetry Maps of Free-Form Curve Segments via Wave Propagation
International Journal of Computer Vision - Special Issue on Computational Vision at Brown University
| Hi-index | 0.00 |
Knowledge about the biological variability of anatomical objects is essential for statistical shape analysis and discrimination between healthy and pathological structures. This paper describes ongoing research on a novel approach that incorporates variability of a training set into the generation of a characteristic 3D shape model. The proposed shape representation is a hybrid of a fine-scale global boundary description and a coarse-scale local medial description. The hybrid overcomes inherent limitations of pure medial based or pure boundary based descriptions. The medial description composed of a net of medial primitives (M-rep) with fixed graph properties is derived from the shape space spanned by the major deformation eigenmodes of a boundary description based on spherical harmonic descriptors (SPHARM). The topology of the M-rep is determined by studying pruned 3D Voronoi skeletons in the given shape space. Its SPHARM descriptors and an individually deformed M-rep model characterize shapes. The hybrid shape description gives an implicit correspondence on the boundary and on the medial manifold, thus enabling a more powerful statistical analysis.