Diabetes is a chronic metabolic disease, which causes, that the body cannot control the blood glucose level, therefor the patient will be in hyperglycemic condition (when the blood sugar level is higher than the normal), and this can lead to many serious issues, like damage of the heart- and vascular system or the nervous system. Diabetes is one of the most widespread diseases, there are more and more people affected by it every year, that’s why the controlling of the blood glucose level is a major research area for the engineers and doctors.
I started my work with a review of the literature of diabetes for the roots and different types of diabetes, the issues it can lead to, and the details of treatment of diabetes with insulin. I also did a research in the field of nutrition and insulin dosing during a day. After that my task was to create a program that simulates the average carbohydrate intake and insulin dosing of a man with diabetes during a day and to test the simulation. The simulated carbohydrate intake plays an important role in the control of diabetes, because with that we can test the control algorithms and compare the results of them to the effect of the manually dosed insulin for the same meals. Based on the listed sources I created the program, which creates the daily carbohydrate and insulin intake vectors. I used these to test more glucose and insulin absorption models, and I represented their results. I also implemented a model of glucose absorption from mixed meals. This is a two compartment model, which keeps track of the amounts of proteins, lipids, sugar and three types of starch in the stomach and of the carbohydrates in the intestine. This model is a good implementation of the realistic nutrition of a man during the day.
My last task was to create an identification method which is able to identify the parameters of glucose and insulin absorption models based on simulated glucose and insulin intake data. I tested the method in three different scenarios, the third one approaches the most the real tasks in which the method can be used. I proved that the method is functioning in these scenarios, and therefore it could be used for parameter identification based on real patient data.