计算机应用

• 人工智能与仿真 •    下一篇

HIC-MedRank基于异构星型网络分析的药物推荐改进算法

邹林霖1,李学明2,李雪3,袁洪4,刘星4   

  1. 1. 重庆大学计算机学院
    2. 重庆大学 计算机学院,重庆 400044
    3. 澳大利亚昆士兰大学信息技术与电子工程学院
    4. 中南大学湘雅三医院心内科
  • 收稿日期:2017-02-07 修回日期:2017-03-15 发布日期:2017-03-15 出版日期:2017-05-13
  • 通讯作者: 李学明

HIC-MedRank: Improved Algorithm for Recommending Drugs by Analyzing Heterogeneous Network

  • Received:2017-02-07 Revised:2017-03-15 Online:2017-03-15 Published:2017-05-13
  • Contact: LI Xueming

摘要: 伴随着医疗文献数据库的快速增长,缺乏经验的初级医师在为患者开处方时难以阅读大量的医疗文献来获得科学地决策辅助。2013年,李雪教授提出MedRank算法,该算法从medline数据库中提取医学信息异构星型网络,基于“有疗效的药物是由好的文章提及的,好的文章是由优秀的作者写的并刊登在高水平的期刊上”的假设,旨在为各类疾病的患者推荐最具有疗效的药物。该算法仍然存在几个问题:第一、模型输入的疾病是复合疾病;第二、推荐的结果不是具体的药物;第三、没有定义判定作者、期刊、文章是“好的”的标准;第四、没有考虑文章的发表时间等其他因素。经过对以上问题的思考,提出了HIC-MedRank算法,该算法纳入作者的H指数、期刊的影响因子、文章的引用数作为评判作者、期刊、文章是否优秀的指标,并综合考虑文章的发表时间、支持机构、发表类型等因素,为高血压合并慢性肾脏病患者推荐最佳的降压药物。推荐的结果与原算法、医师投票的结果以及美国成人高血压治疗指南推荐药物进行对比,实验结果显示HIC-MedRank推荐的药物比原算法推荐的药物更为精准,与主治医师投票选择的药物较为一致,与指南推荐的药物一致性达到80%。

Abstract: Abstract: With the rapid growth of literature,it is difficult for physicians to maintain up-to-date knowledge by reading biomedical literature. An algorithm named MedRank can be used to recommend influential medications from literature by analyzing information network, based on the assumption that “a good treatment is likely to be found in a good medical article published in a good journal, written by good author(s)”. But there are several problems exist in this algorithm: the diseases, as the inputs, are not independent and without complications, the outputs are not specific drugs, and the critical problem is the algorithm didn’t define the criteria on that what kinds of articles, journals and authors are good. This paper propose an improved algorithm named HIC-MedRank, which introduces authors’ H-index, journal’s impact factor and article’s citation counts to evaluate whether an article is good, a journal is excellent, or the authors are eminent. The new algorithm also improved the inputs and outputs, aimed to recommend antihypertensive agents for the patients suffered from Hypertension with Chronic Kidney Disease. The experimental results are compared with the expert rankings collected from physicians and the JNC guidelines. The evaluation shows that the new algorithm’ recommendation drugs are justified with 80% consistency with guidelines’. The HIC-MedRank algorithm can be used to greatly improve the effectiveness of recommendations of influential antihypertensive drugs for chronic kidney disease patients and also can be used to recommend drugs for any other diseases.

中图分类号: