Bayesian Networks for Medical Decision Support Project Abstract:

The treatment of a wide range of medical conditions requires clinicians to recommend appropriate treatment pathways for patients. The selection of which treatment to recommend involves the consideration of patient history and symptoms, test results and quality-of-life issues.

This is a complex task for a number of reasons. Medical data is often imprecise or incomplete due to difficulty of measurement. Medical conditions are multi-factorial with significant and subtle interactions. Data is a mixture of qualitative and quantitative factors. There is often marked variability in the response to treatment from patient to patient for genetic or environmental reasons that may be difficult to identify.

Bayesian Networks (BN) are probabilistic models useful for reasoning with, or representing knowledge under uncertainty. Essentially a BN consists of a Directed Acyclic Graph with nodes representing random variables connected by directed edges representing their (causal) dependencies. The BN gives a factorization of the joint probability distribution (j.p.d.) of the variables. This can be used to support decision making by allowing exploration of the probabilities associated with particular decision paths.