Condition Monitoring in the Adult ICU

The presence of artifact in intensive care monitoring data is a major problem; e.g. maintaining blood pressure in critically ill patients is a key management goal and yet it is the physiological variable most prone to error. In addition to real-time monitoring, artifact detection is necessary for the proper audit or trial of therapies.

This project has developed and validated the Factorial Switching Linear Dynamical System (FSLDS) and the Discriminative Switching Linear Dynamical System (DSLDS) for the detection, removal and cleaning of artifact from vital signs data.

Academic papers

Input-Output Non-Linear Dynamical Systems applied to Physiological Condition Monitoring pdf
Konstantinos Georgatzis, Christopher K.I. Williams and Christopher Hawthorne. Proc Machine Learning in Health Care, JMLR W&C Track Volume 56, 2016.

Detecting Artifactual Events in Vital Signs Monitoring Data. pdf
Partha Lal, Christopher K. I. Williams, Konstantinos Georgatzis, Christopher Hawthorne, Paul McMonagle, Ian Piper, Martin Shaw. Technical report, October 2015. Associated software
A slightly revised version is published as a chapter in Machine Learning for Healthcare Technologies, ed. David A. Clifton, Institution of Engineering and Technology, 2016.

Discriminative Switching Linear Dynamical Systems applied to Physiological Condition Monitoring pdf
Konstantinos Georgatzis, Christopher K. I. Williams. In Proc UAI 2015.
An earlier version was posted on arXiv 24 April 2015 pdf.

Automatic Detection of Artifact and Events in Vital-Signs Traces: A Comparison of Two Machine Learning-Based Models. pdf
Abstract presented at the British Neurosurgical Research Group Meeting in March 2015.

Artifact in physiological data collected from brain injured patients: quantifying the problem and providing a solution through a factorial switching linear dynamical systems approach. pdf
K. Georgatzis, P. Lal, C. Hawthorne, M. Shaw, I. Piper, C. Tarbert, R. Donald, C. K. I. Williams. This initial work on the project was presented orally at the 15th International Symposium on Intracranical Pressure and Brain Monitoring on 6 November 2013; Published in Intracranial Pressure and Brain Monitoring XV, Volume 122 of the series Acta Neurochirurgica Supplement pp 301-305, Springer 2016.


The project was funded by grant number CHZ/4/801 from the Chief Scientist Office (Scotland): Improving Decision Support for Treating Arterial Hypotension in Adult Patients During their Management in Intensive Care, May 2013-Apr 2015. PI: Prof Chris Williams, coIs: Dr Ian Piper, Prof John Kinsella, Prof Peter Andrews, Mr Laurence Dunn. The work of PhD student Konstantinos Georgatzis was supported by SICSA.
Chris Williams