Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.686286
Title: Modelling, optimisation and model predictive control of insulin delivery systems in Type 1 Diabetes Mellitus
Author: Zavitsanou, Stamatina
ISNI:       0000 0004 5918 4131
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2014
Availability of Full Text:
Access through EThOS:
Full text unavailable from EThOS. Please try the link below.
Access through Institution:
Abstract:
Type 1 Diabetes Mellitus is a metabolic disease requiring lifelong treatment with exogenous insulin which significantly affects patient's lifestyle. Therefore, it is of paramount importance to develop novel drug delivery techniques that achieve therapeutic efficacy and ensure patient safety with a minimum impact on their quality of life. Motivated by the challenge to improve the living standard of a diabetic patient, the idea of an artificial pancreas that mimics the endocrine function of a healthy pancreas has been developed in the scientific society. Towards this direction, model predictive control has been established as a very promising control strategy for blood glucose regulation in a system that is dominated by high intra- and inter-patient variability, long time delays, and presence of unknown disturbances such as diet, physical activity and stress levels. This thesis presents a framework for blood glucose regulation with optimal insulin infusion which consists of the following steps: 1. Development of a novel physiologically based compartmental model analysed up to organ level that describes glucose-insulin interactions in type 1 diabetes, 2. Derivation of an approximate model suitable for control applications, 3. Design of an appropriate control strategy and 4. In-silico validation of the closed loop control performance. The developed model's accuracy and prediction ability is evaluated with data obtained from the literature and the UVa/Padova Simulator model, the model parameters are individually estimated and their effect on the model's measured output, the blood glucose concentration, is identified. The model is then linearised and reduced to derive low-order linear approximations of the underlying system suitable for control applications. The proposed control design aims towards an individualised optimal insulin delivery that consists of a patient-specific model predictive controller, a state estimator, a personalised scheduling level and an open loop optimisation problem subjected to patient specific process model and constraints. This control design is modifiable to address the case of limited patient data availability resulting in an 'approximation' control strategy. Both designs are validated in-silico in the presence of predefined, measured and unknown meal disturbances using both the proposed model and the UVa/Padova Simulator model as a virtual patient. The robustness of the control performance is evaluated in several conditions such as skipped meals, variability in the meal content, time and metabolic uncertainty. The simulation results of the closed loop validation studies indicate that the proposed control strategies can achieve promising glycaemic control as demonstrated by the study data. However, further prospective validation of the closed loop control strategy with real patient data is required.
Supervisor: Pistikopoulos, Stratos Sponsor: European Research Council
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.686286  DOI: Not available
Share: