Use this URL to cite or link to this record in EThOS:
Title: Neural networks for medical condition prediction : an investigation of neonatal respiratory disorder
Author: Braithwaite, Emma Annette
Awarding Body: University of Edinburgh
Current Institution: University of Edinburgh
Date of Award: 1998
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This thesis investigates how various signal processing techniques can be applied to diagnose problems in the medical domain. In particular it concentrates on breathing problems often experienced by premature babies who undergo artificial respiration. Medical Decision Support is an area of increasing research interest. The neonatal intensive care unit (NICU) is a prime example. This thesis describes the investigation of techniques to be used as the core of a decision support device in Edinburgh's NICU. At present physiological signals are taken from the patient and archived, little diagnostic use is made of these signals and no investigation has taken place into their diagnostic relevance. Within the scope of the work an investigation has taken place into the application area and some of its current problems have been identified. From these a physiological problem, respiratory disorder, was identified with characteristics which made it worthy of detailed study: it was extremely common, moreover expert knowledge and data about it already existed. With the current techniques the development of respiratory disorder is often missed or diagnosed too late. Signal processing techniques were evaluated with a view to applying them to predict the onset, or classify the development of, respiratory disorder, and a multi-layer perceptron network was chosen to perform as a classifier in the decision support tool. A number of tests were run which included an investigation of the efficiency of the chosen feature extraction techniques and the diagnostic relevance (with respect to the condition under investigation) of the signals being used to assist in diagnosis. Results show that at present the signals of greatest diagnostic relevance are not always used: a decision support device can be developed using a multi-layer perceptron classifier in combination with other signal processing techniques. The thesis also identifies other techniques where there is potential for improving the decision support tool's predictive and classification ability.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available