No abstract

A totally automated rule-based expert system has been developed for interpreting three-dimensional (3D) myocardial perfusion distributions obtained from thallium-201 tomographic images. Over two hundred heuristic rules have been generated for interpreting stress perfusion defects and their characteristics. Perfusion defects were identified in terms of pixels below gender-matched "normal" patient distributions; perfusion defects which reversed with time were identified in terms pixels above gender-matched "normal" patient distributions. The expert system automatically calculates certainty factors from each perfusion defect to provide the relative certainty associated with the location and shape of each myocardial perfusion defect, as well as with the presence, location, and character of each coronary lesion.The current expert system, which was developed after reviewing 294 nuclear tomographic perfusion studies, is being tested and refined using a pilot group of 20 patients.