Reference: Galper, A. R.; Rennels, G. D.; Shortliffe, E. H.; & Patil, R. S. Medicine, AI in. Knowledge Systems Laboratory, April, 1992.
Abstract: This article describes the theoretical basis and current research themes of the field of artificial intelligence in medicine (AIM). Medical advice systems were first envisioned in the late 1950s; more than thirty years later, systems have been built in numerous medical domains, using a wealth of modeling and reasoning techniques. We briefly discuss: the use of protocol analysis in the development of theories of medical problem solving; the nature of medical knowledge; taxonomies of biomedical concepts; knowledge structures; inference and control methods for medical reasoning; uncertainty management; and evaluation functions. Research themes we address include: knowledge acquisition; decision-theoretic approaches; causal reasoning; temporal reasoning and planning; strategies for the diagnosis of multiple diseases; explanation and critiquing; and validation and evaluation of medical expert systems. Finally, the issues and themes we describe are illustrated in the context of several AIM programs--QMR, Pathfinder/Intellipath, and ONCOCIN-- that have achieved limited clinical use.