Blood Glucose Control in Type 1 Diabetes with Deep Reinforcement Learning

What is it about?

Our publication proposes a computer algorithm that uses deep reinforcement learning to control blood glucose levels in patients with type 1 diabetes. We tested the algorithm on a simulation of the glucoregulatory system and found that it performed better than other methods, achieving longer periods of safe glycemic state and lower risk. This could lead to better management of blood glucose levels in type 1 diabetes patients and improve their overall health outcomes.

Featured Image

Photo by Towfiqu barbhuiya on Unsplash

Photo by Towfiqu barbhuiya on Unsplash

Why is it important?

Our work is unique in proposing a deep reinforcement learning algorithm for controlling blood glucose levels in patients with type 1 diabetes. We have tested the algorithm on a simulation of the glucoregulatory system and compared its performance to other well-known alternatives, including a PID controller. We have described our implementation strategy and related problems and compared them with recent proposals using a different implementation strategy. Our results have shown that our strategy can effectively control blood glucose levels, outperforming control baselines in terms of the fraction of time spent in the desired glycemic state and risk metrics. Our work has the potential to improve the management of blood glucose levels in type 1 diabetes patients and improve their overall health outcomes. All the source codes used for evaluation and training as well as the trained agent policies described in this paper are publicly available in our repository.