文章摘要
杜超,范馨月,单立平.利用随机森林建立输尿管上段结石预后预测模型[J].中华医学图书情报杂志,2019,28(5):15-19.
利用随机森林建立输尿管上段结石预后预测模型
Establishment of outcome prediction model for upper ureteral stones using random forest
投稿时间:2019-04-06  
DOI:10.3969/j.issn.1671-3982.2019.05.004
中文关键词: 机器学习  随机森林  输尿管上段结石  预后预测模型
英文关键词: Machine learning  Random forest  Upper ureteral stones  Outcome prediction model
基金项目:
作者单位
杜超 中国医科大学附属盛京医院,辽宁 沈阳 110004 
范馨月 中国医科大学附属盛京医院,辽宁 沈阳 110004 
单立平 中国医科大学附属盛京医院,辽宁 沈阳 110004 
摘要点击次数: 521
全文下载次数: 370
中文摘要:
      目的:采用机器学习方法建立输尿管上段结石患者不同手术方式预后预测模型,为输尿管上段结石的手术选择提供参考。方法:收集2018年1-8月在中国医科大学附属盛京医院泌尿外科接受手术治疗的输尿管上段结石患者的多种临床变量,使用Weka软件根据信息增益率筛选变量,采用SMOTE算法处理数据不平衡问题,利用随机森林方法构建预后模型,并与利用其他3种常见的机器学习算法(NB、SVM、ANNs)得到的模型进行性能比较。结果:结石横面的长径、短径、手术方式等因素在建模中起重要作用,应用随机森林算法构建的输尿管上段结石患者不同手术方式预后预测模型的预测准确率达87.3%,AUC值高达0.902,与其他算法相比效果最佳。结论:基于输尿管上段结石患者的多种临床信息,通过机器学习方法建立的输尿管上段结石预后预测模型能够达到较好效果,在患者术前手术方式的个性化选择上可以为临床医生提供一定的参考。
英文摘要:
      Objective To provide the reference for clinicians to select the surgical procedure for upper ureteral stones by establishing the outcome prediction model of different surgical procedures for upper ureteral stones using machine learning methods. Methods The clinical variables of patients with upper ureteral stones who underwent surgical treatment in our hospital from January 2018 to August 2018 were screened using Weka software according to their information gain rate. The data imbalance was pretreated with SMOTE algorithm. The outcome prediction model of different surgical procedures for upper ureteral stones was established using random forest and its performance was compared with that of the models established using the other 3 machine learning algorithms (NB, SVM and ANNs). Results The long and short diameters of stone cross section and the surgical procedures for upper ureteral stones played an important role in establishing the outcome prediction model of different surgical procedures for upper ureteral stones. The accuracy of the outcome prediction model of different surgical procedures for upper ureteral stones established using random forest was 87.3% and its AUC value was 0.902, which were higher than those of the models established using the other 3 machine learning algorithms. Conclusion The outcome prediction model of different surgical procedures for upper ureteral stones established using random forest based on the clinical information of patients with upper ureteral stones can provide certain reference for the clinicians to select the surgical procedure for upper ureteral stones.
HTML    查看全文   查看/发表评论  下载PDF阅读器
手机扫一扫看