Abstract

A knowledge-based computer system, designed to assist pathologists in the histological diagnosis of breast disease, is described. This system represents knowledge in the form of "disease profiles" and uses a novel inference model based on the mathematical technique of hypergraphs. Its design overcomes many of the limitations of existing expert system technologies when applied to breast disease. In particular, the system can quickly focus on a differential problem and thus reduce the amount of data necessary to reach a conclusion. The system was tested on two sets of samples, consisting of 14 retrospective cases and five hypothetical cases of breast disease. Its recommendations were judged "correct" by the evaluating pathologist in 15 cases. This study shows the feasibility of providing "decision support" in histopathology.