Optimizing bolus calculator sensitivity setting using self-organizing map neural network

Advisors: Reinaldo Moraga; Shi-Jie Chen.Committee members: Christine Nguyen.Includes bibliographical references.Includes illustrations.Type 1 diabetes is an autoimmune disease that affects millions of people around the world. This disease, which has no cure, is treated by injecting insulin into the body. This is done to balance out the body's blood glucose level to help it stay within a healthy range. Blood glucose outside this range can cause many complications and can lead to death if left untreated. In order to maintain a healthy blood glucose level, many technologies have been developed, including a bolus calculator equipped insulin pump and a continuous glucose monitor. The monitor takes blood sugar readings and passes them to the insulin pump, which will notify the user of their current blood glucose level. The user then has the ability to enter the proper amount of insulin required, as determine by the bolus calculator, to lower their current blood glucose level. This calculator is affective for blood glucose levels up to 250 mg/dL, but is insufficient for blood glucose levels above 250 mg/dL due to the body's reaction to the high blood glucose amounts. In order to determine the proper correction bolus amount, this thesis developed a Self-Organizing Map Neural Network to analyze historical blood sugar levels after a correction bolus was complete. In the future, this neural network, which can be implemented alongside bolus calculators, will enable a diabetic to receive a personalized bolus amount recommendation based on their own historic blood glucose values.M.S. (Master of Science