Use this URL to cite or link to this record in EThOS:
Title: Automated multi-parameter monitoring of neonates
Author: Gangadharan, V.
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Advancements in monitoring technology have led to an increasing amount of physiological data; such as heart rate and oxygen saturation, being accumulated in hospitals. A high rate of false alarms in the neonatal intensive care environment due to inadequate analysis of data highlights the need for an intelligent detection system with improved specificity that provides timely alerts to allow early clinical intervention. Current cot-side monitoring systems analyse data channels independently by applying parameter limits. This simplistic system produces a large proportion of false alerts rendering it futile. Ideally, algorithms should be developed to analyse the data to produce clinically useful information and a model that combines data channels for intelligent analysis is desirable. In this thesis, the typical data channels measured in the neonatal intensive care unit (NICU) and the complications encountered in the NICU are reviewed. Previous work on automated patient monitoring and the various approaches for this project are examined. The model would be required to learn ‘stable’ data and indicate novelty when data patterns deviate significantly from this reference. It needs to handle data with unknown underlying distributions. Three statistical models are compared to evaluate their suitability to extract clinical events from neonatal physiological data. These include kernel density estimation (KDE), kernel principal component analysis (KPCA) and one-class support vector machines (OCSVM). The goal is to reduce the false detection rate without altering the sensitivity that is achieved by the current monitor system. Further work involved improving the model and investigating the effect of incorporating additional parameters derived from the data. Physiological data from cot-side monitors and clinical data relating to patient deterioration were collected from nine premature infants. All three novelty detection models that were developed were able to improve event detection from negligible specificity, provided by current methods, to over 70% whilst maintaining high sensitivity.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available