Interval-valued inference in medical knowledge-based system CLINAID

Abstract

In a series of papers and a monograph, we have described the conceptual structures as well as the basic architecture of the knowledge-based system Clinaid. Its generic architecture is aimed at support of knowledge-based decision making with risk and under uncertainty. The majority of extant medical expert systems deal with a limited medical context, the largest domain of knowledge being just a single medical field. The inherent limitation of such medical expert systems is in its essence conceptual and logical: their knowledge bases and inference engines cannot mix easily the knowledge from several fields without some adverse effects. Clinaid deals with this problem by introducing a multi-centre architecture in the Diagnostic Unit. The medical data and knowledge of each medical specialist field exhibit

different logical properties. This in turn leads to the several kinds of many-valued logics on which the relational inference and data manipulation is based.

Semantic and epistemological justification of the use of these logics for the interval-valued inference is provided by a theoretical device called the checklist paradigm. The Diagnostic unit of Clinaid deals with a number of body systems. In this paper we shall use several body systems to demonstrate the interaction across the body systems during the inferential process and to show how interval based methods of inference help in dealing with the problems induced by this phenomenon.