A general purpose shell for research assessment of Bayesian knowledge bases supporting medical diagnostic software systems

The principle of separately identifying an inference engine and a knowledge base, commonly used in AI expert systems, has been realized in a C program that utilizes a Bayesian inference engine easily used in research and clinical applications. Given a knowledge base file of relative weights for the input features of a group of medical conditions, a user can enter features of individual cases interactively or in either of two streamlined modes. A case can be saved in or retrieved from a file, and the disease probabilities calculated from the knowledge base can be displayed on the screen at any time. Finally, a batch mode permits rapid processing of large numbers of cases whose features are contained in a single disk file. The diagnostic probabilities for each case are written to a collection output file. This Bayesian shell permits rapid evaluation of Bayesian knowledge bases in support of knowledge base and program development. It can be embedded in other programs that utilize Bayes' theorem to evaluate any body of data, clinical or otherwise.<>