A knowledge representation methodology for the development of knowledge-based systems for the detection and diagnosis of cancer

Current artificial intelligence research includes expanding the problem domain while maintaining power capabilities. There exist expert systems which detect or diagnose specific types of cancer. Ideally, this domain could be expanded to excompass a wider range of cancer types. Similarities in different cancers were documented in an attempt to find a knowledge representation methodology which would support domain expansion. The observed similarities in the cancer staging systems were compared with the knowledge representation methodologies used in the implementation of some current cancer expert systems. It was then determined that domain expansion could best be achieved by devoloping a new knowledge representation methodology which supports the standard staging systems used in cancer diagnosis. The new cancer representation methodology is a frame-based semantic network developed specifically for the cancer domain although future applications are widespread. The system uses four types of links and four types of entities developed to take advantage of the similarities in the different cancer types. The links and entities meld together to form a knowledge representation methodology suitable for a range of cancers. The new knowledge representation methodology exhibits the desirable properties of a frame-based semantic network and some new considerations which have not yet been readily implemented in other knowledge-based systems. The system demonstrates improvements over existing systems which approach similar problems and represents a significant step in the construction of a knowledge-based expert system for a group of cancers.