Abstract

In the present paper some formal aspects of the hypothesis-directed stage of medical diagnosis are studied and an algorithm of the diagnostic problem solving process is described. A given field of medical knowledge is represented by a pair of graphs. The sentences describing observed symptoms and signs constitute the data on which the algorithm is based. In the first step, the set of true judgments is determined and the hypotheses which are impossible in a given situation are rejected. In consequence the model of knowledge is modified and working hypotheses are chosen. Further steps consist in selection of hypotheses in view of testing. The set of expected symptoms is determined and these are classified according to their diagnostic value. The process ends when the conditions of arriving at a useful solution of a given problem are fulfilled. The description of the algorithm is based on a clinical example. The model aims at reflecting some of the most important structural features of the diagnosis and does not embrace the probability evaluation problems.