Key result
Reinforcement learning improves long-term adaptation of automated insulin delivery vs rule-based controllers in-silico.
Why the study?
Most automated insulin delivery systems are evaluated only for short-term performance, risking inadequate control if not adapted for variations in physiology affecting insulin sensitivity.
Do reinforcement learning agents improve the long-term adaptation of automated insulin delivery systems compared to rule-based approaches in in-silico adult models?
Do reinforcement learning agents improve the long-term adaptation of automated insulin delivery systems compared to rule-based approaches in in-silico adult models?
Reinforcement learning agents using continuous reward functions improve the long-term adaptation of automated insulin delivery systems in in-silico models.
Answers are grounded in this paper and cite their sources.
Controller adaptation for IS variations may enhance AID performance; leaves open long-term clinical translation in prospective trials.
Several closed or hybrid loop controllers for Blood Glucose (BG) regulation, which are also known as Artificial Pancreas (AP) Systems or Automated Insulin Delivery systems (AIDs), are in development worldwide. Most AIDs are designed and evaluated for short-term performance, with a particular emphasis on the post-meal period. However, if controllers are not adapted properly to account for variations in physiology that affect Insulin Sensitivity (IS), the AIDs may perform inadequately. In this work, the performance of two Reinforcement Learning (RL) agents trained under both piecewise and continuous reward functions is evaluated in-silico for long-term adaptation of a Fully Automated Insulin Delivery (fAID) system. An automatic adaptive discretization scheme that expands the state space as needed is also implemented to avoid disproportionate state space exploration. The proposed agents are evaluated for long-term adaptation of the Automatic Regulation of Glucose (ARG) algorithm, considering variations in IS. Results show that both RL agents have improved performance compared to a rule-based decision-making approach and the baseline controller for the majority of the adult population. Moreover, the use of a continuous shaped reward function proves to enhance the performance of the agents further than a piecewise one.
No takes yet. Share an insight, caveat, or question.
Serafini et al. (2023) studied Blood Glucose regulation. Reinforcement Learning (RL) agents trained under piecewise and continuous reward functions vs. Rule-based decision-making approach and baseline controller was evaluated on Performance for long-term adaptation of the Automatic Regulation of Glucose (ARG) algorithm. Reinforcement Learning agents trained with continuous reward functions improved long-term adaptation of a fully automated insulin delivery system compared to rule-based and baseline controllers in-silico.
Synapse has enriched 10 related papers. Consider them for comparative context: