Analysis of biomedical images using maximal matching

Many biomedical images share the property that their objects and object properties are highly predictable. This makes model-matching techniques attractive for recognition of those objects. Here we examine some ways for generating and testing correspondences between "observed" entities and objects of the model that may be useful for diagnosis. In particular, a method is given to efficiently generate a sequence of maximal matchings, and an initial experiment applying a matching technique to chest x-ray data is described.