EVIN Diego Alexis
Temporal Abstraction for the Analysis of Intensive Care Information
ALEJANDRO HADAD; DIEGO A. EVIN; BARTOLOME DROZDOWICZ; OMAR CHIOTTI
Journal of Physics: Conference Series
Lugar: Philadelphia; Año: 2007 vol. 90 p. 1 - 1
This paper proposes a scheme for the analysis of time-stamped series data from multiple monitoring devices of intensive care units, using Temporal Abstraction concepts. Thisscheme is oriented to obtain a description of the patient state evolution in an unsupervised way.The case of study is based on a dataset clinically classified with Pulmonary Edema. For thisdataset a trends based Temporal Abstraction mechanism is proposed, by means of aBehaviours Base of time-stamped series and then used in a classification step. Combining this approach with the introduction of expert knowledge, using Fuzzy Logic, and multivariate analysis by means of Self-Organizing Maps, a states characterization model is obtained. This model is feasible of being extended to different patients groups and states. The proposed scheme allows to obtain intermediate states descriptions through which it is passing the patient and that could be used to anticipate alert situations.