An Intelligent Alarm System Based on Temporal Episodes for Intensive Care Units
| Authors: | Charbonnier Sylvie, INPG, France, Metropolitan
Badji Lyes, INPG, France, Metropolitan
Gentil Sylviane, INPG, France, Metropolitan |
|---|
| Topic: | 8.2 Modelling & Control of Biomedical Systems |
|---|
| Session: | Control Application to Life Medicine II |
|---|
| Keywords: | Biomedical engineering, monitoring elements, alarm filtering, knowledge acquisition, trends analysis, qualitative signal representation |
|---|
Abstract
This paper presents an alarm system based on temporal episodes. A temporal episode is a semi qualitative information that describes the signal trend. The system developed is a limit alarm system based on episodes extracted on line. It is able to recognize specific signal behaviors such as probe disconnection or steady state near the alarm limit threshold. The system is implemented to run on line and was tested on 36 hours of data recorded on adult patients hospitalized in Intensive Care Units. The alarm periods generated are compared to those raised by a classical limit alarm system. The conclusion is that 35% of the alarms raised by the classical system would be filtered, without any clinically relevant alarm missed.
