A low-cost personal computer-based radiology diagnostic expert system and image and text database

A program for radiology differential diagnosis (DDX) using Prolog, a programming language for artificial intelligence applications is developed. The program uses a probabilistic model based on Bayes´ theorem to arrive at and rank the relative probabilities of possible diagnosis. This model uses the prevalence of diagnoses and conditional probabilities of diagnostic features. It treats nonindependent diagnostic features by defining compound conditional probabilities to describe the interdependence of conditional probabilities. The program has an iterative mouse-controlled menu-driven interface that makes selecting more diagnostic features yield shorter DDX´s. This process can be reversed by deselecting features, allowing what-if experimentation. The program ranks the DDX by relative probability and allows reviewing images or text descriptions of each diagnosis in the list. The program creates separate databases that users may alter, using terminology of their choosing for diagnoses and diagnostic features