| Intelligent, automatic methods of monitoring large groups of patients in hospitals and in their homes are urgently required. This presentation describes machine learning techniques for building personalised models of human physiological behaviour, leading to intelligent patient monitoring systems that are used to improve the care of hospital patients and patients in their homes. Our research is funded by the UK Government as a priority centre in personalised medicine and patient safety.
Growth in demand for intelligent patient monitoring systems is being driven by the ever-increasing complexity and volume of data acquired from patients. Patient deaths can be avoided if sufficiently intelligent monitoring approaches are available.
We will describe how artificial intelligence and statistics can be used to provide early warning of patient deterioration. These techniques have been demonstrated both in aircraft engine monitoring and in the monitoring of acutely ill hospital patients. Our system is being used throughout the whole patient journey in the hospital, from the arrival in the ambulance through to discharge home.
Our patient monitoring data-fusion system is based on a probabilistic model of patient data learned from one of the world’s largest datasets of vital signs acquired from high-risk hospital patients. This system alerts the nursing staff whenever the combination of vital sign parameters is indicative of physiological abnormality, and has undergone clinical trials in UK and US hospitals. It has reduced three-fold the percentage of patients requiring emergency care, from 18% to 5%. |
