A critical review of trend-detection methodologies for biomedical monitoring systems.

Abstract Information gained through early detection of patient vital sign changes typically can be used to anticipate future difficulties. Detection of these changes through monitoring, however, can be difficult. Many of the monitored processes are random in nature. For that reason simple threshold algorithms exhibit a high incidence of false alarms, which can decrease the operator's confidence in the monitor. Many problems associated with threshold-based biological monitors can be alleviated by introducing statistical detection techniques. The purpose of this review is to develop and critique the major trend detection algorithms used in biological monitors. This review contains an evolution of trend detection through current state-of-the-art algorithms.