An Introduction to Active Shape Models

Biomedical images usually contain complex objects, which will vary in appearance significantly from one image to another. Attempting to measure or detect the presence of particular structures in such images can be a daunting task. The inherent variability will thwart naive schemes. However, by using models which can cope with the variability it is possible to successfully analyse complex images. Here we will consider a number of methods where the model represents the expected shape and local grey-level structure of a target object in an image. Model-based methods make use of a prior model of what is expected in the image, and typically attempt to find the best match of the model to the data in a new image. Having matched the model, one can then make measurements or test whether the target is actually present. This approach is a ‘top-down’ strategy, and differs significantly from ‘bottomup’ methods. In the latter the image data is examined at a low level, looking for local structures such as edges or regions, which are assembled into groups in an attempt to identify objects of interest. Without a global model of what to expect, this approach is difficult and prone to failure. A wide variety of model based approaches have been explored (see the review below). This chapter will concentrate on a statistical approach, in which a model is built from analysing the appearance of a set of labelled examples. Where structures vary in shape or texture, it is possible to learn what are plausible variations and what are not. A new image can be interpreted by finding the best plausible match of the model to the image data. The advantages of such a method are that