Learning Shape Models from Examples Using Automatic Shape Clustering and Procrustes Analysis

A new fully automated shape learning method is presented. It is based on clustering a shape training set in the original shape space and performing a Procrustes analysis on each cluster to obtain a cluster prototype and information about shape variation. As a direct application of our shape learning method, a 17-structure shape model of brain substructures was computed from MR image data, an eigen-shape model was automatically trained, and employed in our method for segmenta-tion of those MR brain images not present in the shape-training set. Our approach can serve as a fully valid automated substitute to the tedious and time-consuming manual shape analysis. 1 1 Motivation Automated learning of shape models is an important problem in medical image analysis with direct implications in the area of medical image interpretation. We and others have previously demonstrated the utility of incorporating shape in medical image segmentation and interpretation 1]. However, training a shape-based segmentation system is mostly done manually following a tedious and therefore impractical process. In the work reported here, a novel approach to automated learning of shape models from examples is presented and its utility demonstrated in segmentation of MR brain images. We have developed a novel solution to the problem of shape reparameterization{ alignment{averaging problem. The main diierence from previously reported methods 2, 3] is that the training set is rst automatically clustered and those shapes considered to be outliers are discarded. The second diierence is in the manner in which registered sets of points are extracted from each shape contour.