冠心病患者痰热互结证CHAID决策树识别模式的研究

目的:建立冠心病不稳定性心绞痛患者临床常规检测指标对痰热互结证的识别模式。方法:选取2010年4月至2011年4月多中心收集的冠心病不稳定性心绞痛患者411例的基本资料、中医四诊信息及临床常规检测指标进行归一化处理后,采用CHAID决策树方法从90个临床常规检测指标中自动提取痰热互结证的识别规律。对其中212例患者进行痰热互结证识别模式的外验证。结果:Cl离子、缩短分数、RDW-CV、血常规RBC、D-Ⅱ聚体、CK-MB、PTA和BUN共8个属性指标经筛选后进入决策树识别模型。该模型对411例患者的测试结果显示:敏感度为75.0%,特异度为86.9%,检验准确率为86.1%。外验证模型缺失RDW-CV,模型识别准确率为85.8%,敏感度为89.5%,特异度为85.5%。结论:临床常规检测指标经CHAID决策树方法筛选后,可以直观、清晰的进行冠心病不稳定性心绞痛患者痰热互结证的识别,自动归纳识别规律,在中医证型-生物学指标对应模式的数据挖掘中具备一定的优势。

Study on Recognition Modes of Phlegm-Heat Stasis Syndrome Decision Tree of Patients with Coronary Heart Diseases

To build a recognition mode of phlegm-heat stasis syndrome according to indexes of clinical routine examinations of patients with unstable angina (UA).Methods:From April 2010 to April 2011,the clinical basic data,TCM four-diagnosis information and clinical routine examination indexes of 411 patients with UA were collected and normalized,and the recognition rules of phlegm-heat stasis syndrome was automatically extracted from 90 clinical examination indexes by applying CHAID decision tree method.The verification of recognition mode of phlegm-heat stasis syndrome in 212 patients was conducted.Results:A total of 8 indicators,including Cl ion,shortening fraction,RDW-CV,blood RBC,D-II mer,CK-MB,PTA and BUN were indexed into the decision tree model.The test results of 411 patients of the model showed that the sensitivity was 75.0%,the specificity was 86.9%,and the correct rate was 86.1%.The verification of association patterns lacked RDW-CV,and its correct rate was 85.8%.The sensitivity was 89.5%,and the specificity was 85.5%.Conclusion:The CHAID decision tree model can clearly and directly conduct recognition of phlegm-heat stasis syndrome with the basis of clinical routine examination indexes,and automatically summarize recognition rules,which has certain advantages in data mining of syndrome-physiochemistry model.