ROLE OF EXPERT SYSTEMS IN CLINICAL DIAGNOSIS.

This article provides an in-depth look at one of the areas most amenable to expert systems research and development. It begins with a discussion of how expert systems can be designed, paying special attention to the problem of medical knowledge representation. It then turns directly to diagnostic problem-solving and identifies the following as critical to medical diagnosis problem-solving: the accumulationa and integration of initial findings; the generation and evaluation of hypotheses; the setting of goals; the evaluation and selection of relevant sources of information; the sorting of findings according to hypotheses and goals; and the integration of evidence. It is concluded that smart systems can be of enormous help to relatively inexperienced diagnosticians but somewhat less so to highly experienced ones. The article then present the MEDAS and RDAS systems, two expert clinical diagnostic systems. Both systems use a modified Bayesian approach to generate probabilities about the likelihood of various diagnoses or various diagnostic outcomes, such as 'positive tendency', 'negative tendency', 'ambiguous', 'verified', 'eliminated', or 'inactive'.