P. Lucas - The Model-based Approach to Medical Decision Support
Medicine is a semantically rich area in which uncertainty plays a major role. Probabilistic graphical models, such as Bayesian networks, offer a versatile and expressive way to model the uncertainty involved in medical decision making. These models also form the basis of almost all the AI oriented research in medical decision making. In this lecture, medical tasks such as diagnosis, prognosis, treatment selection and follow-up are cast probabilistic and decision-theoretic problems. The mapping of particular characteristics of the medical domain under consideration to specific probabilistic constructs is also discussed. A number of medical dicision-making models developed in collaboration with clinicians are discussed in detail to illustrate general principles.