POMDP在住院2型糖尿病治疗方案优化研究中的应用  

Application of POMDP in the Optimization of Hospitalized Type 2 Diabetes Treatment Program

作　　者：杨帆[1] 李海英[1] 许少杰[1] 邹兆重[1] 李忆琴[1] 陈画虹[1] YANG Fan LI Haiying XU Shaojie ZOU Zhaozhong LI Yiqin CHEN Huahong(Boai Hospital of Zhongshan, Zhongshan, Guangdong 528403, Chin)

机构地区：[1]中山市博爱医院,广东中山528403

出　　处：《今日药学》2017年第9期632-634,648,共4页Pharmacy Today

基　　金：中山市科技计划项目(2015B1180)

摘　　要：目的探讨部分可观察马尔可夫决策过程(POMDP)在住院2型糖尿病治疗方案优化研究中的应用问题。方法回顾性分析某院2012~2016年住院治疗的650例2型糖尿病患者的相关生化指标及临床用药数据等资料。采用MATLAB R2014a进行POMDP挖掘分析,以临床最大效应值的水平为标准,评价不同治疗方案的临床疗效。结果采用Gephi0.8.2筛选出治疗方案核心用药。POMDP模型评价了2型糖尿病临床治疗方案,发现优化的核心治疗方案。结论 POMDP模型可以从临床数据中挖掘出优化的治疗方案。OBJECTIVE To explore the application of Partially Observable Markov Decision Process（ POMDP） in the optimization research of hospitalized type 2 diabetes treatment programs. METHODS Data of 650 type 2 diabetes mellitus patients hospitalized from 2012 to 2016 was collected,biochemical indicators and clinical drug use data were analyzed. POMDP Mining Analysis was performed on MATLAB R2014 a, clinical efficacy of different treatment programs regimens were evaluated and compared.RESULTS Used Gephi0.8.2 to filter out the core medication of the treatment programs. POMDP Model identified the core treatment programs for type 2 diabetes. CONCLUSION POMDP Model can be applied to discover the optimized core treatment programs from the clinical data.

关 键 词：部分可观察马尔可夫决策过程 住院2型糖尿病 治疗方案 优化 

分 类 号：R969.3[医药卫生—药理学]