Adaptive and Predictive Alarm Management in Intensive Care Units


Sahiika Genc, Ph.D.

Software Sciences  and Analytics Lab

General Electric Global Research

Friday,  February 3, 2012

3:30 – 4:30p.m

Rm. 1500 EECS


Abstract: This talk is composed of two parts: 1) Automatic adaptive alarm limit and level management and 2) model-based predictive hypotension alarm for Intensive Care Units (ICUs). In the first part, we describe an automatic adaptive alarm limit and level management logic, to reduce workflow for electrocardiographic monitoring devices, based on a framework utilizing a collection of concurrent and asynchronous Timed Input/Output Automata. The key purpose of limit and level adaptation is to reduce False Positive (FP) alarms (without increasing False Negatives - FN), boosting Sensitivity and positive predictive value (PPV), while the key purpose of limit and level management is to reduce workflow of physicians, monitoring technicians, nurses, and aides. Performance results are provided on a data set consisting of 158 alarm records, roughly approximating a month's admissions in an Intensive Care Unit at a large hospital. In the second part of the talk, we discuss a stochastic model of the cardiovascular system that predicts the evolution of the mean arterial pressure (MAP) and provides a stochastic prediction metric for acute hypotensive episodes. We consider the complete arterial baroreceptor reflex pathway as a closed-loop control system composed of an effector portiobaroreceptors, their afferent nerve fibers, the medullary cardiovascular centers (MCCs), and efferent sympathetic and parasympathetic fibers. The performance of the model is tested on a case study of acute hypotensive episodes (AHEs) on PhysioNet data.

Biosketch: Sahika Genc received her B.S. degree in Electrical and Electronics Engineering from the Middle East Technical University, Ankara, Turkey in 2000, and Ph.D. degree in Electrical Engineering: Systems from the University of Michigan, Ann Arbor, Michigan in 2006. She joined the General Electric Global Research in Niskayuna, New York in August 2006. She is currently located in the Software Sciences and Analytics Lab. Her current research interests include alarm management and clinical decision support algorithms for physiological systems, and control and optimization for power distribution networks. She holds 11 patent applications and a trade secret. Sahika enjoys teaching. Since Fall 2010, she has been serving as an adjunct faculty at the Department of Electrical, Computer, & Systems Engineering in Rensselaer Polytechnic Institute, Troy, NY, where she teaches senior level control systems courses.