Diagnosing Lyme disease - Tailoring patient specific Bayesian networks for temporal reasoning.
Kristian G. Olesen, Ole Kristian Hejlesen, Ram Benny Dessau, Ivan Beltoft, Michael Trangeled · VBN Forskningsportal (Aalborg Universitet) · 2006
Lyme disease is an infection evolving in three stages. Lyme disease is characterised by a number of symptoms whose manifestations evolve over time. In order to correctly classify the disease it is important to include the clinical history of the patient. Consultations are typically scattered at non-equidistant points in time and the probability of observing symptoms depend on the time since the disease was inflicted on the patient. A simple model of the evolution of symptoms over time forms the basis of a dynamically tailored model that describes a specific patient. The time of infliction of the disease is estimated by a model search that identifies the most probable model for the patient given the pattern of symptom manifestations over time. 1