REPRESENTATION OF SYMPTOMS AND EVIDENCE IN DIAGNOSIS SUPPORT Suitability of a diagnosis support and its intuitive transparency depend on medical knowledge

Suitability of a diagnosis support and its intuitive transparency depend on medical knowledge representation. An appropriate representation should solve the following problems: a consistent interpretation of symptoms of different nature, formulation of rules that is close to diagnostic heuristics and implementation of a credibility measure that estimates diagnostic hypotheses. The paper suggests a description of symptoms by means of classical and fuzzy sets and a model of diagnostic reasoning in the framework of the Dempster-Shafer theory. Thus, credibility of the diagnosis is defined by the imprecision measure (i.e. a characteristic or a membership function) and by the certainty of the rule (i.e. a basic probability assignment value). A unified approach is proposed for all kinds of symptoms and evidence that are used during the diagnosis. Models for symptoms and evidence that are difficult to define (e.g. pain), are suggested.