万海玉,李 军,王 博,等.应用贝叶斯网络建立孤立性肺结节良恶性预测模型[J].肿瘤学杂志,2022,28(5):380-384.
应用贝叶斯网络建立孤立性肺结节良恶性预测模型
Establishment of Prediction Model for Isolated Pulmonary Benign or Malignant Nodule by Bayesian Network
投稿时间:2021-11-10  
DOI:10.11735/j.issn.1671-170X.2022.05.B007
中文关键词:  肺结节  贝叶斯网络  诊断
英文关键词:pulmonary nodule  Bayesian network  diagnosis
基金项目:
作者单位
万海玉 山西医科大学第一临床医学院 
李 军 山西白求恩医院 
王 博 山西医科大学第一临床医学院 
张小艳 山西医科大学第一临床医学院 
摘要点击次数: 642
全文下载次数: 200
中文摘要:
      摘 要:[目的] 探讨基于贝叶斯网络建立孤立性肺结节良恶性预测模型的临床价值。[方法] 从医院数据管理平台收集2018年12月至2021年10月山西医科大学第一医院576例孤立性肺结节患者的人口学资料、临床资料及影像学资料。根据患者的病理结果分为良恶性两组,其中458例肺癌,118例肺良性疾病。分析采用 SPSS 26.0 软件进行卡方检验,并进行共线性诊断分析,进行模型变量的初步筛选。使用Clementine12.0软件进行TAN贝叶斯网络模型的建立。[结果] 单因素初步分析显示肺癌与肺良性疾病组间年龄、呼吸系统疾病既往史、个人恶性肿瘤史、结节直径、部位、钙化、分叶征、胸膜凹陷、血管集束征、边缘和结节的密度均有统计学差异(P<0.05)。将上述11个变量输入建立TAN贝叶斯网络,其准确率为91.27%、灵敏度为92.23%、特异度为86.96%、阳性预测值为96.94%、阴性预测值为71.43%。[结论] 应用贝叶斯网络构建的肺结节良恶性的预测模型具有较好的预测能力,并且能够更加直观地描述疾病与因素间复杂的网络风险机制。
英文摘要:
      Abstract:[Objective] To establish a prediction model for isolated pulmonary benign or malignant nodules based on Bayesian network. [Methods] The demographic data, clinical data and imaging data of 576 patients with isolated pulmonary nodule in the First Hospital of Shanxi Medical University from December 2018 to October 2021 were collected from hospital data management platform. Patients were divided into benign group(n=118) and malignant group(n=458) according to pathological result. Chi-square test was performed with SPSS 26.0 software for analysis, and colinear diagnostic analysis was performed for preliminary screening of model variables. Clementine 12.0 software was used to establish the TAN Bayesian network model. [Results] There were significant differences in age, respiratory disease history, personal history of malignant tumor, diameter and location of nodule, calcification, lobulation sign, pleural depression, vascular cluster sign, margin and density of nodule between benign and malignant group(all P<0.05). The above 11 variables were input to establish TAN Bayesian network. The accuracy, sensitivity, specificity positive predictive value and negative predictive value of the prediction model were 91.27%, 92.23%, 86.96%, 96.94% and 71.43%, respectively. [Conclusion] The prediction model for differentiating benign and malignant pulmonary nodule based on Bayesian network has good prediction ability and can describe more intuitively the complex network risk mechanisms between disease and factors.
在线阅读   查看全文  查看/发表评论  下载PDF阅读器