Transatlantic transferability of a new reinforcement learning model for optimizing haemodynamic treatment for critically ill patients with sepsis

作者：

Highlights：

• Deep Q reinforcement model for the treatment of sepsis could provide potential clinical decision support.

• Developed policy is transferable across a different patient population.

• Novel Deep policy inspection method that provides valuable interpretable insight into policy behavior.

• Compared to current clinical practice, the developed policy tends to prefer a more liberal vasopressor therapy and more restrictive fluid resuscitation.

摘要

•Deep Q reinforcement model for the treatment of sepsis could provide potential clinical decision support.•Developed policy is transferable across a different patient population.•Novel Deep policy inspection method that provides valuable interpretable insight into policy behavior.•Compared to current clinical practice, the developed policy tends to prefer a more liberal vasopressor therapy and more restrictive fluid resuscitation.

论文关键词：Sepsis,Reinforcement learning,Deep Q learning,ICU

论文评审过程：Received 14 February 2020, Revised 16 November 2020, Accepted 8 December 2020, Available online 15 December 2020, Version of Record 3 February 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2020.102003