An Expert System Developed from a Hard Data Knowledge Base: Example of a Laboratory Based Anemia Consultant

摘要

An expert system derived from a hard data knowledge base has been developed for the laboratory diagnosis of anemias. Five hundred sixty patient records were reviewed and analyzed in a knowledge base system, the CONSULT LEARNING SYSTEM (CLS). Probabilities obtained from the CLS were used to train the expert system, CONSULT-I. This paper is a preliminary report of the development of an expert system from hard data (patient records), the modification of this training by expert hematologists, and performance evaluation.CONSULT-I is a system of artificial intelligence and statistical pattern recognition for developing expert classification systems. Multiple presentations of categories may be stored in its knowledge base. In this study, only two presentations were stored: (1) hard data alone obtained from the 560 patient records, and (2) the hard data presentation edited by experts using rules of Insignificance and Can't.The utility of extending a low cost screening test (CBC with smear evaluation) by means of an expert system is appealing for potential cost reduction in health care. The expert system allows for immediate use of the multi-dimensional knowledge stored in 30 features about the CBC, and provides Actions oriented toward more immediate diagnosis.