冠心病病人血瘀证的决策树模型研究

Study on Decision Tree Patterns of Coronary Heart Disease Patients with Blood Stasis Syndrome

目的建立基于临床四诊信息的冠心病不稳定型心绞痛(UA)病人血瘀证的诊断模式,探索与UA病人血瘀证相关的四诊信息的组合模式。方法对411例UA病人进行临床证候学调查,获得包括症状、舌象、脉象在内的共计174项临床四诊信息。在t检验分析基础上,分别采用C4.5、ADTree、CHAID及QUEST决策树方法进行数据分析,构建能够区分UA病人血瘀证的诊断模式。结果 411例UA病人血瘀证比例占58.9%。采用t检验分析得到具有统计学意义的四诊信息30项,按照四诊信息的显著性顺序依次进入C4.5、ADTree、CHAID及QUEST决策树模型,分别筛选出能够区分血瘀证的最优模型的诊断准确率分别为:76.7%、75.7%、73.7%和69.6%。10倍交叉验证结果显示C4.5、ADTree及CHAID决策树模型优于QUEST模型。结论决策树模型可以基于临床四诊信息数据清晰、直观的进行UA病人血瘀证的识别,归纳诊断规律。

Objective To establish a diagnosis pattern of unstable angina （UA） patients with blood stasis based on four clinical di- agnostic information and explore the combination pattern of four diagnostic information. Methods Clinical syndromes survey was taken to obtain 174 four diagnostic information including symptoms, tongue pictures, and pulse. C4.5, ADTree, CHAID, and QUEST methods were applied to analysis the data and establish the diagnosis pattern of UA patients with blood stasis syndrome. Results Among the 411 patients,242 were with blood stasis syndrome （58.9 %）. Thirty four diagnostic information,which had significant differences after T - test analysis, were turned into decision tree patter. The accuracy of the most four optimal patterns reached 76.7 %, 75.7 %, 73.7 % and 69.6 %. Ten- fold cross- validation results showed that C4.5, ADTree,and CHAID decision tree patterns were superior to QUEST. Conclusion, Decision tree pattern could identify blood stasis syndrome of UA patients clearly and more intuitive and gen- eralize diagnosis rules based on the data of clinical four diagnostic information.