Near-optimal insulin treatment for diabetes patients: A machine learning approach
作者：
Highlights：
• We present a diabetes control based on a stochastic model of blood glucose level (BGL).
• We introduce the health reward function (HRF) which numerically grades a BGL.
• HRF and the stochastic model give rise to a Markov Decision Process (MDP).
• The solution to MDP provides an optimal policy which leads to healthy simulated BGL.
• The method is equally effective for non-standard and rare diabetic patients.
摘要
•We present a diabetes control based on a stochastic model of blood glucose level (BGL).•We introduce the health reward function (HRF) which numerically grades a BGL.•HRF and the stochastic model give rise to a Markov Decision Process (MDP).•The solution to MDP provides an optimal policy which leads to healthy simulated BGL.•The method is equally effective for non-standard and rare diabetic patients.
论文关键词：
论文评审过程：Received 15 July 2019, Revised 17 June 2020, Accepted 23 June 2020, Available online 29 June 2020, Version of Record 24 July 2020.
论文官网地址：https://doi.org/10.1016/j.artmed.2020.101917
