Beta 1


Title Detection of Events using Continuous Glucose Monitors
Author Pejtersen, Mette
Jensen, Karen Sandø
Supervisor Jørgensen, John Bagterp (Scientific Computing, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Bachelor thesis
Year 2010
Abstract In this B.Sc. thesis a bolus calculator is created to estimate the correct amount of insulin to deliver in relation to a meal, and an algorithm is developed that can automatically detect forgotten bolus and catheter drop-out events. These two events are some of the most common types of treatment problems for people with type 1 diabetes, and if a patient can be warned early when they occur, then he or she can take corrective action and thereby achieve a more tightly regulated glucose concentration. The purpose of the bolus calculator and the algorithm is to improve the quality of life of, and minimize the risk of long-term diabetes related complications for people suffering from type 1 diabetes. First, a simple bolus calculator is investigated. It takes the current glucose level, the amount of carbohydrates (CHO) in a meal, the amount of active insulin in the body, the insulin-to-CHO ratio and the insulin sensitivity factor into account. The insulin-to-CHO ratio and the insulin sensitivity facter are estimated for one virtual subject. For days with three normal meals and a correct estimation of CHO, use of the simple bolus calculator to estimate the boluses helps to keep the glucose concentration at an acceptable level. The bolus calculator underestimates the boluses for meals less than 250 g CHO and overestimates the bolus for meals greater than 250 g CHO. A correction bolus is necessary when a bolus in connection with a meal is underestimated. The calculation of the correction bolus takes the remaining active insulin in the body into account. Subsequently, an event detecting algorithm is developed that can detect forgotten bolus and insulin catheter drop-out events automatically. The algorithm first detects increases in the glucose level. When there is an increase beyond a certain threshold, four criteria have to be fulfilled to ensure that there is an event and not just a normal meal-bolus situation or a noise peak. The parameters in the criteria are optimized to get as few false alarms as possible and to detect the events as fast as possible to avoid hyperglycemia. The algorithm is then tested on a virtual subject simulated using the Hovorka Model, a physiological model of the glucose-insulin interaction in type 1 diabetes. The subject is simulated for four different month-long periods. The meals are based on normal meal sizes and times that are randomized to simulate realistic conditions as closely as possible. The corresponding doses of insulin are calculated with the bolus calculator. There are eight abnormal events, four forgotten bolus events and four catheter drop-out events, that are simulated over the course of the simulations. The algorithm detects all the events. The number of false alarms is kept reasonably low. Most of the false alarms occur during fasting periods, either during the night or during fasting days, and all the false alarms are caused by noise. The events are detected at acceptable mean glucose concentrations between 8-9 mmol/L.
Imprint Technical University of Denmark (DTU) : Kgs. Lyngby, Denmark
Series IMM-B.Sc.-2010-10
Fulltext
Original PDF bac10_10.pdf (1.88 MB)
Admin Creation date: 2010-07-02    Update date: 2010-07-02    Source: dtu    ID: 264598    Original MXD