New Models for Expert System Design

Aitken, James Stuart (1990) New Models for Expert System Design. PhD thesis, University of Glasgow.

Full text available as:
[img]
Preview
PDF
Download (4MB) | Preview

Abstract

This thesis presents new work on the analysis of human lung sound. Experimental studies investigated the relationship between the condition of the lungs and the power spectrum of lung sound detected at the chest wall. The conclusion drawn from two clinical studies was that the median frequency of the lung sound power spectrum increases with a decrease in airway calibre. The technique for the analysis of lung sound presented in this thesis is a non-invasive method which may be capable of assessing differences in airway calibre between different lobes of the lung. An expert system for the analysis of lung sound data and pulmonary function data was designed. The expert knowledge was expressed in a belief logic, a system of logic which is more expressive than first order logic. New automated theorem proving methods were developed for the belief logic. The new methods were implemented to form the 'inference engine' of the expert system. The new expert system compared favourably with systems which perform a similar task. The use of belief logic allows introspective reasoning to be carried out. Plausible reasoning, a type of introspective reasoning which allows conclusions to be drawn when the database is incomplete, was proposed and tested. The author concludes that the use of a belief logic in expert system design has significant advantages over conventional approaches. The experimental results of the lung sound research were incorporated into the expert system rule base: the medical and expert system research were complementary.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Electrical engineering, Biomedical engineering
Date of Award: 1990
Depositing User: Enlighten Team
Unique ID: glathesis:1990-78073
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 28 Feb 2020 12:09
Last Modified: 28 Feb 2020 12:09
URI: http://theses.gla.ac.uk/id/eprint/78073

Actions (login required)

View Item View Item