Soft-computing control methods for type 1 diabetes

OData support
Supervisor:
Dr. Kovács Levente
Department of Control Engineering and Information Technology

Diabetes is ranked to the epidemic category by the World Health Organization (WHO), thus the research of this topic has high priority in the 21st century.

The main topic of my thesis was the automatic control of the blood-sugar level of the insulin-dependent (Diabetes type I.) patients. I designed control algorithms based on soft computing techniques for the Intensive Care Unit model (ICU) of the glucose-insulin dynamics, which is one of the several models published in the last five years.

In order to approximate the system optimally in noisy measuring conditions I made Kalman-filter for four sophisticated models (Lotz, Wong, Suhaimi and ICING). After this filter design an Adaptive Neuro-Fuzzy Interference System (ANFIS) was implemented. The combination of the original neural and a fuzzy system inherited only the favorable characteristics of the two types of controllers.

The completed universal approximator provides the right conclusion even if the experimental data are imprecise. Moreover this system is able to learn from the output-input pairs of the dynamic model.

The last step of my work was to embed these controllers to the Diabetes Simulator designed by Medical Informatics Laboratory of the Budapest University of Technology and Economics (BUTE).

Downloads

Please sign in to download the files of this thesis.