Bayesian networks for disease diagnosis: What are they, who has used them and how?

A Bayesian network (BN) is a probabilistic graph based on Bayes' theorem, used to show dependencies or cause-and-effect relationships between variables. They are widely applied in diagnostic processes since they allow the incorporation of medical knowledge to the model while expressing uncertainty in terms of probability. This systematic review presents the state of the art in the applications of BNs in medicine in general and in the diagnosis and prognosis of diseases in particular. Indexed articles from the last 40 years were included. The studies generally used the typical measures of diagnostic and prognostic accuracy: sensitivity, specificity, accuracy, precision, and the area under the ROC curve. Overall, we found that disease diagnosis and prognosis based on BNs can be successfully used to model complex medical problems that require reasoning under conditions of uncertainty.