Automating hospital ICU emergency signaling

Abstract

This thesis verifies that the proposed Kernel mapping based recursive least square algorithm can detect the slightest deviation of anomaly from the norm, monitor and learn underlying pattern between natural and abnormal multivariate medical parameters of a particular critical ICU patient with high detection accuracy and very low rate of false alarm. This online, automated, sequential, real-time intruder detection algorithm is suitable for any instantaneous detection of accidental emergencies without compromising the patient safety and effectiveness of care. It is an elegant, inexpensive solution, independent of complexity, and also a portable and adaptive approach.