ADA: An online trend pattern detection system

Pattern recognition has been used extensively in medical information retrieval and data analyses. Specifically, it involves pattern classification, indexing, clustering, anomaly detection and rule detection. Among various patterns, trend is a simple yet powerful pattern that can be associated with many complex clinical symptoms. Detecting adverse clinical trend is thus an important proactive approach to critical clinical situation managements. In this paper, we propose an online trend pattern detection system, the Anaesthetic Data Analyser (ADA), as a platform to monitor trend patterns of physiological data collected during anaesthesia. ADA differentiates from current approaches by looking at trends rather than a single data value against a preset threshold. Our online trend pattern detection and trend query processing algorithms also make ADA support real time trend monitoring efficiently. Experiments on physiological data collected from patients demonstrate the efficiency and effectiveness of the ADA system and our algorithms