Enabling real-time intelligent decision support in intensive care

Abstract

Medical devices in ICU allow for both continuous monitoring of patients and data collection. Nevertheless, the amount of data to be considered is such that it is difficult for doctors to extract all the useful knowledge. In order to help uncover some of that knowledge we have built an IDSS based in the agent's paradigm and using data mining techniques to build prediction models. With the intention of collecting as much data as possible the data acquisition process was automated. Furthermore, given the paramount importance of data quality for data mining a data quality agent responsible for detecting the errors in the data was devised. Indeed, data acquisition in the ICU is error prone as, for instance, sensors may be displaced as patients move. The aim of this paper is to present the overall KDD process implemented, presenting in detail the data transformations that were done and the benefits achieved.