A temporal data mining approach for discovering knowledge on the changes of the patient's physiology

Abstract

Physiological data represent the health conditions of a patient over time. They can be analyzed to gain knowledge on the course of a disease or, more generally, on the physiology of a patient. Typical approaches rely on background medical knowledge to track or recognize single stages of the disease. However, when no one domain knowledge is available these approaches become inapplicable. In this paper we describe a Temporal Data Mining approach to acquire knowledge about the possible causes which can trigger particular stages of the disease or, more generally, which can determine changes in the patient's physiology. The analysis is performed in two steps: first, identification of the states of the disease (namely, the stages through which the physiology evolves), then detection of the events which may determine the change from a state to the next one. Computational solutions to both issues are presented. The application to the scenario of the sleep disorders allows to discover events, in the form of breathing and cardiovascular disorders, which may trigger particular sleep stages. Results are evaluated and discussed. © 2009 Springer Berlin Heidelberg.