基于贝叶斯网络构建慢性阻塞性肺疾病急性加重期痰热壅肺证症状与证候间关联模式

Constructing Symptom and Syndrome Association Mode of Acute Exacerbation of Chronic Obstructive Pulmonary Disease with Phlegm-heat Congesting Lung Syndrome: Based on Bayesian Network

目的基于贝叶斯网络探讨慢性阻塞性肺疾病急性加重期症状对证候的贡献度,构建症状与痰热壅肺证的关联模式。方法采用全国九家省级中医院为期2年所调查的1768例患者数据建立痰热壅肺证数据库。数据预处理后导入贝叶斯网络。以痰热壅肺证为目标变量,以决策树运行结果筛选出的变量为输入变量,运用IBM SPSS Modeler 14.2软件实现贝叶斯网络算法,选用评估及分析对TAN及马尔科夫毯两种模型所获得的结果进行比较分析。结果痰热壅肺证因果关系如下:1)痰色黄、痰白黏、脉滑数、脉数、大便秘结、口渴喜冷饮;2)痰色黄、痰白黏、脉滑数、舌苔黄腻;3)痰色黄、痰白黏、脉滑数、舌苔黄、舌苔黄腻;4)痰色黄、痰白黏、脉滑数、舌质红。上述4条中,共有的症状为痰色黄、痰白黏、脉滑数,其中,痰色黄对痰热壅肺证的贡献度最大,为0.72;其次痰白黏,为0.61。马尔科夫毯模型显示,口渴喜冷饮、大便秘结、舌苔黄腻对痰热壅肺证的贡献度均>0.7。结论痰热壅肺证与症状间有4条因果关系,痰色黄、口渴喜冷饮、大便秘结、舌苔黄腻对痰热壅肺证的贡献度较大。

Objective To explore the contribution of symptoms to syndromes of chronic obstructive pulmonary disease acute exacerbation based on Bayesian network and construct the association mode of symptoms and phlegmheat congesting lung syndrome. Methods Phlegm-heat congesting lung syndrome database was built adopting 1768 patients＇ data researched in 9 national provincial Chinese medicine hospitals for 2 years. Data were imported in Bayesian network after preprocessing. Taking phlegm-heat congesting lung syndrome as the target variable,taking the variables selected from the results of the decision tree as the input variables,Bayesian network algorithm was carried out by IBM SPSS Modeler 14. 2 software. Comparing analysis was made on the results gained from TAN and markov blanket models using assessment and analysis. Results Phlegm-heat congesting lung syndrome causality showed below.First,Yellow phlegm,white sticky phlegm,slippery-rapid pulse,rapid pulse,constipation,thirst and liking cold drink. Second,Yellow phlegm,white sticky phlegm,slippery-rapid pulse and yellow greasy fur. Third,Yellow phlegm,white sticky phlegm,slippery-rapid pulse,yellow fur and yellow greasy fur. Forth,Yellow phlegm,white sticky phlegm,slippery-rapid pulse and red tongue. In the above 4 items,the common symptoms were yellow phlegm,white sticky phlegm and slippery-rapid pulse. Among them,yellow phlegm made the most contribution,0. 72. White sticky phlegm was the second,0. 61. Markov blanket model showed that the contribution of thirst and liking cold drink,constipation and yellow greasy fur to phlegm-heat congesting lung syndrome was more than 0. 7.Conclusion There were 4 items of causality between phlegm-heat congesting lung syndrome and symptoms. Yellow phlegm,thirst and liking cold drink,constipation and yellow greasy fur made great contributions to phlegm-heat congesting lung syndrome.