Scope of the project is to develop and evaluate in clinical setup a mobile Health (mHealth) based platform for the optimization of the insulin treatment in individuals with type 1 diabetes (T1D). The platform will be comprised of a glucose monitoring device, a highly accurate patch pump and a novel algorithmic approach. The algorithm will be based on a reinforcement learning approach for the personalization of insulin treatment. The use of an algorithm based on a self-learning approach, which adapts to the physiological and behavioural characteristics of each patient, will allow the enhanced glycemic control of individuals with diabetes.
Characters of glucoregulatory system
- Complex and non-linear
- Described by limited information
- Characterized by inter- and intra-patient variability
- Influenced by a number of factors e.g. meals, exercise (disturbances)
- Characterized by delays
Characters of reinforcement learning
Reinforcement learning (RL), branch of machine learning (ML), is an intensively active research field which embraces algorithms able to learn from data and perform optimization within uncertain environments. The field of RL falls between supervised and unsupervised learning and includes problems where an agent attempts to improve its performance over time at a given task through the continual interaction with its environment. RL began to develop as an independent branch in the early 1980s inspired by the animal psychology and the idea of learning through trial-and-error. It was quickly adopted by the field of optimal control as a very efficient way to solve dynamic programming problems for which the Bellman’s “curse of dimensionality” restricted an analytic solution. RL is a field with an extensively investigated theoretical background, which now finds its way towards practical application with the modern advancement in computational capacity. In this view, the application in real-life problems is being highlighted as one of the current trends of RL.
MyTreat
ABBA: Adaptive Basal-Bolus Algorithm
