Designing an experimental expert system for lung tuberculosis diagnostics using fuzzy set theory

This paper describes a diagnosis system for tuberculosis called TUBERDIAG which is able to combine positive and negative knowledge. The knowledge base of the system consists of a set of IF-THEN rules for diagnosis of tuberculosis where each rule assigns its weight in [0,1]. The inference machine will produce the final diagnosis as a total degree of confirmation and exclusion of tuberculosis diagnosis. This paper focuses on a description of the structure of TUBERDIAG. The system is developed in cooperation with doctors at the National Institute of Tuberculosis and Lung Disease, Hanoi. The system is written in Borland Delphi 2.0 programming language for Windows 95 and can run on PCs.