Chapter 3 - Correspondence Establishment in Statistical Shape Modeling: Optimization and Evaluation

Abstract

Statistical shape modeling has been widely used in medical-imaging applications, where the shape and possible shape variation of an anatomic structure are derived from a set of shape instances of the structure. Shape correspondence, which aims to accurately identify corresponding landmarks across the given set of shape instances, is one of the main challenges in constructing a statistical shape model. In this chapter, we will discuss several important research topics in shape correspondence, including 1) pairwise shape correspondence, 2) groupwise shape correspondence, and 3) shape-correspondence performance evaluation. For simplicity, we focus on 2D shape correspondence in this chapter, with a brief discussion of the extensions to 3D shape correspondence at the end. Experiments on sample structures are presented to evaluate the performance of the proposed methods.