天津科技大学学报2025，Vol.40Issue(2)：71-80,10.DOI:10.13364/j.issn.1672-6510.20230203
基于强化学习框架的脓毒症抗生素多策略推荐模型
A Multi-Policy Recommendation Model for Antibiotic Use in Sepsis Based on Reinforcement Learning Framework
摘要
Abstract
Sepsis is one of the leading causes of human death worldwide,and antibiotics are an important part of sepsis treatment.In recent years,researchers have considered medical decision-making problems as Markov decision processes and used reinforcement learning methods for treatment strategy recommendations.In this article we propose a multi-policy rec-ommendation framework combining value-based and policy-based reinforcement learning methods for antibiotic use in sep-sis treatment.Different decision regions are defined based on patient characteristic information,and the multi-policy model provides personalized treatment recommendations.The results show that our multi-policy selection model can achieve a good prognosis for patients in 80.32%of cases.Through statistical analysis of decision trajectories and drug action selection,our model can provide reasonable drug recommendations in accordance with clinical practice,and recommend appropriate anti-biotic combinations to improve patient prognosis.关键词
强化学习/医疗决策/脓毒症Key words
reinforcement learning/medical decision-making/sepsis分类
信息技术与安全科学引用本文复制引用
王嫄,刘安岐,盛梦茹,侯佳佳,赵婷婷,于琦..基于强化学习框架的脓毒症抗生素多策略推荐模型[J].天津科技大学学报,2025,40(2):71-80,10.基金项目
国家自然科学基金项目(61976156) （61976156）
天津市科技特派员项目(20YDTPJC00560) （20YDTPJC00560）
