Title: MONITORING SYSTEMS IN INTENSIVE CARE UNITS USING INCREMENTAL SUPPORT VECTOR MACHINES

Year of Publication: 2013
Page Numbers: 273-280
Authors: Fahmi Ben Rejab, KaoutherNouira, Abdelwahed Trabelsi
Conference Name: The Third International Conference on Digital Information Processing and Communications (ICDIPC2013)
- United Arab Emirates

Abstract:


In this paper, we adapt the incremental Support Vector Machines, denoted by I-SVM to the medical field in order to solve the issue of monitoring system in Intensive Care Unit (ICU). In fact, we have already proved the effectiveness of the SVM to reduce false alarms in batch mode in ICU. In this work, our purpose is to prove that the SVM can provide spectacular results not only in batch modes but also in online mode in training phase. In fact, the main problem in ICU consists of the increasing of the rate of false alarms what presents a real threat for the patient life. Our work aims to decrease this rate and to monitor the different states of patients by using the I-SVM. This technique will allow us to incorporate new measurements relative to patients through its ability to handle the large amount of data streams in ICU. Besides, it will provide more correct description for the patient states. As a result, the I-SVM will generate more accurate alarms than the current system. Experiments using real-medical databases have shown its efficiency to reduce the rate of false alarms and to reproduce true ones compared to both the current monitoring system and the system using SVM on batch mode.