Electronic health records based reinforcement learning for treatment optimizing

作者:

Highlights:

• We propose an EHRs based reinforcement learning method to optimize the treatment suggestions of diseases that requires sequential decision making.

• We conduct blood glucose control tasks on DKA patients with our model. The results show the rationality and superiority of the proposed method.

• Inspired by multi-agent scenarios, we use cooperative learning with linear value decomposition to improve the performances of benchmark model.

摘要

•We propose an EHRs based reinforcement learning method to optimize the treatment suggestions of diseases that requires sequential decision making.•We conduct blood glucose control tasks on DKA patients with our model. The results show the rationality and superiority of the proposed method.•Inspired by multi-agent scenarios, we use cooperative learning with linear value decomposition to improve the performances of benchmark model.

论文关键词:Electronic health records,Deep reinforcement learning,Glucose control,Cooperative learning

论文评审过程:Received 27 April 2021, Revised 1 July 2021, Accepted 26 August 2021, Available online 10 September 2021, Version of Record 23 September 2021.

论文官网地址:https://doi.org/10.1016/j.is.2021.101878