This research proposes to develop a semantic image database that could serve investigations in cellular pathology within three sub domains: (1) Cancer, the storage/detection of tumor/cellular images; (2) Cellular biology, requiring the identification of nuclei in different stages of cellular division; and (3) Neurobiology, quantification of cell types and neuropathological objects in human or animal brain tissue.
The proposed research advances and integrates two basic information technologies: iterative image object classification and object relational knowledge base. The former is employed to interact with domain experts to define models of object and concept, and the latter is employed to retrieve image information based on such models in a declarative fashion. Based on such, the research is aimed to develop a complete set of key concepts to support semantic biological image retrieval in neurobiology and cancer research.
Although semantic retrieval of arbitrary images is, in general, a very difficult problem, specific domains of biological images may be sufficiently constrained in contents that one might hope to achieve semantic retrieval. The success of the proposed research will be a demonstration of this principle and lead to the development of many other special purpose semantic image systems. Finally, the proposed research will introduce new requirements to the areas of object relational database, decision support system and data mining system.