Use of neural network techniques in a medical expert system

Abstract

Expert systems in medicine have relied heavily upon knowledge-based techniques, in which decision making rules or strategies are derived through consultation with experts. These techniques, coupled with methods of approximate reasoning, have produced systems which model the human decision making process. This approach has the disadvantage of requiring extensive interviewing of experts for each new application. It is desirable to be able to supplement this information by extracting information directly from data bases, without expert intervention. In this article, a neural network model is used to extract this information, and then use it in conjunction with rule-based knowledge, incorporating techniques of approximate reasoning.