Representing uncertainties in a diagnostic expert system for coronary arterial stenosis

A medical diagnostic system was constructed using the multiplication model of support-logical programming in order to determine the predominant stenosis in one of the three main coronary arteries using scintigrams. Stress thallium-201 planar scintigrams for the patients in the study were categorized by coronary arteriography; 66 patients had >70% stenosis of a major epicardial artery and 25 patients had no arterial stenosis. Support logic programming was selected as the technique to represent the uncertainties in the interpretation of the scintigrams. The reasoning process of cardiologists was successfully modeled, specifically for diagnosing of the site of the predominant coronary arterial stenosis. A Prolog-based expert system was constructed from information obtained through consultation with a nuclear cardiologist. The diagnostic concepts were modeled in the form of production rules. Dominance of a stenosis in each of the three main coronary arteries was simultaneously considered. With the current set of production rules, the system properly recognized the site of coronary arterial stenosis in 64 of the 66 patients presented.<>