Consistency checking of binary categorical relationships in a medical knowledge base

摘要

Knowledge bases of medical expert systems have grown to such an extent that formal methods to verify their consistency seem highly desirable; otherwise, decision results of such expert systems are not reliable and contradictory entries in the knowledge base may cause erroneous conclusions.This paper presents a new formalization of the finding/finding, finding/disease, and disease/disease relationships of the medical expert system Cadiag-1. This formalization also helps to clarify the differences between the application of propositional logic and of quantificational logic to capture the meaning of some fundamental categorical relationships in the area of medical diagnostics. Moreover, this formalization leads to very simple yet provably correct and complete algorithms to check the consistency of a medical knowledge base containing a set of these relationships.