Preparing Medical Knowledge for Diagnostic Expert Systems

Abstract

The task of supporting medical diagnosis by computer applications is twofold: first, we have to accumulate and integrate medical knowledge into some kind of information system; secondly, we can apply these findings to actual patient data by generating and evaluating patient-specific hypotheses for diagnostics. This paper is focused on the initial task for computer-supported medical diagnosis. We present an approach to organize and later on, to represent medical knowledge in an information system, regardless to the original diagnostic problem domain.

Due to the knowledge we already have about medical decision making, the organization of medical knowledge can be predefined to a certain extent: Diagnostics requires several entrypoints to previously accumulated, and thus, case-independent medical knowledge: diseases, symptoms, factors like etiology, morphology, etc . Moreover, medical decision making reqnires not only the static coverage of the problem domain but also the representation of diagnostic procedures to support its dynamic aspects.

In order to meet these extraordinary requirements for the representation of medical knowledge we applied a comprehensive modelling technique, namely object-oriented design . It allows us not only to cover static properties as well as basic processes of medical diagnostics appropriately, but also to implement such a complex application accurately. For the representation of additional diagnostic heuristics and the design of working hypotheses by the diagnostician we have extended the representation schema by rules . Thus, for implementation a hybrid system, namely Prolog-DB (a data base coupled with a logical programming language) had to be used.