Deep Reinforcement Learning for Fractionated Radiotherapy in Non-Small Cell Lung Carcinoma

作者：

Highlights：

• Development of a DRL-based controller for personalized radiation therapy treatment;

• Development of a virtual radiotherapy environment for the DRL training;

• Parameter estimation by Particle Swarm Optimisation and routinely collected CT scans;

• Use of three reward functions modelling different treatments' aggressiveness;

• DRL approach can adapt to radiotherapy outperforming the current clinical practice.

摘要

•Development of a DRL-based controller for personalized radiation therapy treatment;•Development of a virtual radiotherapy environment for the DRL training;•Parameter estimation by Particle Swarm Optimisation and routinely collected CT scans;•Use of three reward functions modelling different treatments' aggressiveness;•DRL approach can adapt to radiotherapy outperforming the current clinical practice.

论文关键词：Deep reinforcement learning,D3QN,Particle swarm optimization,NSCLC,Radiation therapy,Tumour treatment optimization

论文评审过程：Received 8 September 2020, Revised 28 May 2021, Accepted 3 August 2021, Available online 15 August 2021, Version of Record 20 August 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102137