A Personalized Algorithm To Control Blood Glucose Levels During Exercise In Individuals With Type 1 Diabetes

"Exercise has numerous well-established benefits, such as decreased risk of cardiovascular disease, improved lipid profile, and overall improved well being. These benefits are especially important to patients with type 1 diabetes, given the increased risk of cardiovascular disease in this population. Despite the established benefits of exercise, moderate intensity aerobic exercise increases the risk of hypoglycemia in individuals with type 1 diabetes, making exercise more difficult in this population. For exercise management in type 1 diabetes, carbohydrate ingestion and insulin reduction are recommended to prevent hypoglycemia. However, due to the large inter-individual variability in glucose responses to exercise, these general recommendations are not always efficient in preventing hypoglycemia. In the present thesis, a personalized closed-loop algorithm based on each patient's glucose response to exercise was developed to reduce the risk of exercise-induced hypoglycemia. The designed algorithm is based on a prediction mathematical model and uses an optimization-based method. After each exercise session, the prediction model is updated by estimating the exercise effect using a least squares algorithm. Given the updated model, an optimization problem is formulated to obtain recommendations of basal rate reduction and carbohydrate intake for the upcoming exercise session. The developed algorithm was evaluated on 100 virtual patients in a computer simulation environment. The results showed that there was a significant reduction in hypoglycemia with the developed algorithm in comparison to the conventional exercise management strategy, without significant increase in time in hyperglycemia. Furthermore, it was shown that when exercise is announced earlier, the algorithm performs better and leads to lower risk of hypoglycemia. The developed algorithm has the potential to facilitate physical activity in type 1 diabetes and thus improve quality of life. Clinical studies to assess the algorithm are warranted"--