A User-Friendly Development Tool for Medical Diagnosis Based on Bayesian Networks

This paper presents a web supported development tool specific for medical diagnosis, based on Bayesian networks. The proposed system provides a user-friendly interface, giving the users (experts in the medical domain) the possibility to design diagnostic applications without deep background knowledge on Bayesian networks theory. Efficient statistical inference mechanisms are implemented, taking advantage of the simple structure of medical diagnostic models, composed of causal disease-symptom relations. The developed system also provides an Internet page interface to interact with the expert, when using the designed application for diagnosis. Illustrating the functionality of the proposed tool, an application in the context of sleep-disorders diagnosis is presented.