A Content Based Pattern Analysis System for a Biological Specimen Collection

Over the years many research collections of biological specimen have been developed for research in biological sciences. Number of specimens in some of these collections can be as high as several millions. There is a move to convert these physical specimens into digital images. This research is motivated by the need to develop techniques to mine useful information from these large collections of specimen images. Specific focus of this research is on the collection of parasites in the Harold W. Manter Laboratory (HWML) Parasite Collection, one of the top four parasite collections in the world. These parasites closely resemble in shape and have flexible bodies with rigid extremities. They have only a few specific structural differences. In this paper we present a technique to retrieve specimens based on shape of a given sample. This form of mining based on the shape of the specimen has the potential to discover linkages between specimens not otherwise known.