Artificial intelligence-Bayesian analysis system for cardiac catheterization laboratory

Clinical decision-support systems for the cardiologist emphasize Bayesian analysis but provide no record-keeping capability or retrospective patient analysis. This study was begun to investigate and develop an efficient method of combining artificial intelligence (AI) techniques with conventional Bayesian analysis, to design and implement an AI expert system for medical record-keeping and consultation in the cardiac catheterization laboratory that provides the capability for retrospective analysis of a patient´s clinical course, and to evaluate the system with actual patient records. A normalized point-score system to predict the probability of coronary artery disease based on noninvasive indices was developed. An expert system, including a natural-language interface and an information-management database, was implemented using the point-score system. Fifteen patients who had received cardiac catheterization were studied with the system. It was found that the expert system was much more ´conservative´ with its recommendations for clinical treatment than human clinical experts. Analysis of the results indicates that it would have been more cost-effective to follow the expert system´s clinical recommendations for 26% of the patients studied.<>