Condition Monitoring in Premature Babies

The project is concerned with detecting patterns in the monitoring traces of premature babies in intensive care, using probabilistic modelling. The goal is to identify different types of artifact and pathology in real time based on characteristic patterns in the data. Currently there are a high number of false alarms in neonatal units due to events like a baby being handled, or a monitoring probe becoming dislodged. By identifying such artifactal factors automatically the aim is to reduce false alarms, and thus to focus attention on the state of health of the baby.

The project is a collaboration between Prof Neil McIntosh (Child Life and Health, University of Edinburgh), Dr Yvonne Freer (Simpson Centre for Reproductive Health, The Royal Infirmary of Edinburgh), and Prof Chris Williams (School of Informatics, University of Edinburgh). The initial work was carried out by John Quinn, and was supported by a grant from the premature baby charity BLISS. The research is being further developed by Ioan Stanculescu.

Academic papers

A Hierarchical Switching Linear Dynamical System Applied to the Detection of Sepsis in Neonatal Condition Monitoring pdf
Ioan Stanculescu, Christopher K. I. Williams, Yvonne Freer
Proceedings of the 30th Conference on Uncertainty in Artificial Intelligence (UAI 2014).

Autoregressive Hidden Markov Models for the Early Detection of Neonatal Sepsis preprint pdf
Ioan Stanculescu, Christopher K.I. Williams, and Yvonne Freer
Accepted for publication in IEEE Journal of Biomedical and Health Informatics, 1 Dec 2013. Published in J-BHI 18(5) 1560-1570, September 2014.

Automating the Calibration of a Neonatal Condition Monitoring System pdf
C.K.I. Williams and I. Stanculescu.
In Proc AIME 2011, eds M. Peleg, N. Lavrač, and C. Combi, LNAI 6747, pp. 240--249. Springer (2011)

Physiological Monitoring with Factorial Switching Linear Dynamical Systems pdf
J.A. Quinn and C.K.I. Williams.
Chapter appearing in Bayesian Time Series Models, eds. D. Barber, T. Cemgil, S. Chiappa, Cambridge University Press, 2011.

Factorial Switching Linear Dynamical Systems applied to Physiological Condition Monitoring pdf
John A. Quinn, Christopher K.I. Williams, Neil McIntosh.
Accepted to IEEE Trans. on Pattern Analysis and Machine Intelligence (July 2008), published T-PAMI 31(9) pp 1537-1551 (2009). Matlab code is available.

Known Unknowns: Novelty Detection in Condition Monitoring pdf
John A. Quinn, Christopher K. I. Williams.
Invited paper in Proc 3rd Iberian Conference on Pattern Recognition and Image Analysis (IbPRIA 2007),
eds J. Marti, J. M. Benedi, A. M. Mendonca, J. Serrat, LNCS 4477 pp 1-6, Springer-Verlag (2007).

Factorial Switching Kalman Filters for Condition Monitoring in Neonatal Intensive Care pdf
Christopher K. I. Williams, John Quinn, Neil McIntosh
In Advances in Neural Information Processing Systems 18, eds. Y. Weiss, B. Schoelkopf, J. C. Platt, MIT Press (2006)

Media Coverage

Chris Williams