AbstractIn a situation of uncertain diagnosis, physicians may spend valuable time consulting relevant literature, often with unsatisfactory results. Therefore, our aim was to develop a computer system which supports differential diagnostics via rapid and comprehensive searches through information in literature. Based on entered signs and symptoms our prototype is able to offer probable diagnoses. Subsequently, further examinations and tests are suggested to confirm or exclude a certain disease. Thus, the final diagnosis is made gradually by differentiation of possible diseases.Our first attempt consisted in representing knowledge in a rule-based PROLOG system. However, because nearly all information for a sign-oriented differential diagnosis can be represented with very few relations only, we turned to a fact-oriented representation of signs and diseases. Access was possible via PROLOG or an imperative programming language. The index-sequential access on a fact-oriented representation of knowledge was suitable to manage a large knowledge base, which is necessary for a thorough differential diagnosis. A pointer structure was recently examined in order to handle different object-to-object relations. Efficient information processing is now possible which provides short response times using even broad knowledge bases.
Integrating Bayesian networks and decision trees in a sequential rule-based transportation model
