IGUANA: Individuation of Global Unsafe ANomalies and Alarm activation

Abstract

In this paper, we present the IGUANA (individuation of global unsafe anomalies and alarm activation) framework which performs analysis of clinical data to characterize the risk level of a patient and identify dangerous situations. Data mining techniques are exploited to build a model of both normal and unsafe situations, which can be tailored to specific behaviors of a given patient clinical situation. A risk function has been proposed to identify the instantaneous risk of each physiological parameter. The classification phase, performed on-line, assigns a risk label to each measured value. We have developed a prototype of IGUANA in R, an open source environment for statistical analyses and graphical visualization, to validate our approach. Experimental results, performed on 64 records of patients affected by different diseases, show the adaptability and the efficiency of the proposed approach