Thesis

Mathematical modelling of blood glucose for short-term diabetes therapy (using artificial neural networks)

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2012
Thesis identifier
  • T13226
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • Diabetes is a major global health issue. Proper control of the blood glucose level (BGL) is very important. Insulin type, dose, stress levels, diet, exercise, height to weight ratio, metabolism rate, etc. all affect the BGL. Sometimes, inconsistency in measurement of the BGL values by the patient or the inability to interpret the data obtained from the system can lead to problems in the management of diabetes. It is therefore necessary to design a system which can assist patients in managing their diabetes. The aim of this project was to design an improved artificial neural network (ANN) which will predict the short-term BGL and improve diet, exercise and insulin regimens. It should also enable training to be continuously updated with each new dataset. An important advantage of using ANNs is that they do not require a comprehensive overview of the problem, but are trained to recognise the patterns in the input dataset which are stored effectively. In this project, an ANN was modelled using the Neural Network ToolboxTM in the MATLABTM software. The automated insulin dosage advisor (AIDA) is a simple model of the interaction of glucose and insulin in the body and is mainly intended for Type-1 diabetes. The data obtained from the AIDA software was used to train the ANN in order to predict the BGL. The R2011b version of MATLABTM was used in this project. Functions such as 'adaptwb' and 'trainlm' were used to update the weights while training. The architecture used was a slightly modified version of the Elman architecture and it was found that the best results obtained were at 82 epochs. However, the effects of the individual inputs given to the network were still unknown.
Resource Type
DOI
Date Created
  • 2012
Former identifier
  • 947889

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T13226 2021-07-02 Public Download