Intelligent Diabetes Assistant: A Telemedicine System for Modeling and Managing Blood Glucose

Abstract

The creation of a diabetes management assistant that can remotely collect data, increase communication between patient and care provider, and automatically analyze all available information could improve the health of many diabetics. Individual models, taking into account nutrition, medication, and exercise, with appropriate mathematical modeling, can learn accurate representations of specific patients suitable for providing therapy advice. The fundamental goal of effective diabetes management is for the patient to select behaviors that maintain glycemic homeostasis. Thus the goal of an intelligent diabetes assistant is to help the patient select optimal behaviors. To do this the assistant must be able to learn how a patient's choices will affect blood glucose. From the care providers perspective a system should be able to provide detailed and accurate data about the patient, increase interaction between patient and expert, and be efficient. This thesis describes an intelligent diabetes assistant (IDA) designed to meet these goals. IDA uses a mobile phone application and other devices to measure the three primary components that affect blood glucose: meals, medication, and exercise. The data are used to learn models for predicting how behaviors around meal times affect postprandial blood glucose, and to create a new continuous physiological model that includes exercise. These models can then be used in a variety of ways to generate therapy advice for the patient and health care provider. The complete system is presented in this thesis.