Characterising obstructive sleep apnea patients through complex networks

Abstract Obstructive sleep apnea is a condition whose evolution is poorly understood and difficult to predict, in spite of its high prevalence and serious complications, due to the complexity of its initial symptoms and systemic consequences. In this contribution we discuss the characterisation of a group of patients suffering from this condition through the use of complex networks. Similarity relationships between different subjects are mapped into a network using the recently proposed convergence/divergence formalism. Topological features are then extracted from this structure, and used to feed a classification model forecasting the future evolution of patients after a standard treatment. Results indicate that the complex network approach is able to extract information over and above standard data mining models, thus yielding a new way for the characterisation, and hence for the understanding, of this complex condition.