Current Methods in Statistical Shape Analysis

Abstract

The extraction of information out of 2D or 3D images often relies on the detection, recognition and interpretation of shapes and shape variabilities. This directly involves the (mathematical) representation of shapes as well as methods to account for and measure the morphological differences. Even though in clinical routine shape analysis is frequently done by viewing the images alone, there is a wide range of applications where automatical methods with formalized metrics are needed for overall data interpretation and shape statistics. This chapter is dedicated to the description of these methods and is divided as follows: First, the importance of shape modeling in medical image analysis is outlined and the concept of statistical shape models and their representations are discussed in section 2.1. Following that, we expand on the fundamental problem of determining correspondences between shapes and on several methods of solution (section 2.2) which directly leads us to discuss the associated statistical shape models in section 2.3. Section 2.4 explores the benefits of statistical shape models for medical image segmentation and discusses explicitly and implicitly represented shape priors.