【摘要】 目的：通过分析疑似冠心病患者中冠心病组与非冠心病组的脉图特征，并基于脉图特征建立冠心病高危人群筛查模型，挖掘脉诊的临床诊断价值。方法：将疑似冠心病患者分为冠心病组与非冠心病组，采用SmartTCM-I型脉象仪采集脉象样本，运用非参数检验统计方法比较两组脉图特征差异；采用随机森林模式识别算法，建立冠心病高危人群筛查模型。结果：与非冠心病组比较，冠心病组时域特征h2/h1、h3/h1显著增大（P<0.01）；多尺度熵MSE1、MSE2、MSE3、MSE4、MSE5显著减小（P<0.05），当脉图时域特征和多尺度熵特征都参入建模时，模型性能最高，模型的准确度最高（95.67%）、特异度最高（93.99%）、误判率最低（6.01%）。结论：脉图特征一定程度上可以反映疑似冠心病患者病理状态及血管功能，基于脉图特征等信息建立的冠心病高危人群筛查模型具有一定的参考意义。更多还原

【Abstract】 Objective: By analyzing TCM pulse characteristics of suspected patients with coronary heart disease(CHD)and establishing the screening model of CHD high-risk population based on the pulse characteristics, the clinical diagnostic value of TCM was explore. Methods: The pulse samples were collected by Smart TCM-I digital pulse acquisition analyzer, which were divided into CHD group and non-CHD group, and the differences of characteristic parameters between the two groups were compared and analyzed by non-parametric test. Random forest pattern recognition algorithm was used to establish a screening model for high-risk population of CHD. Results: Compared with non-CHD group, the time-domain characteristics h2/h1 and h3/h1 in CHD group pulse were significantly increased(P<0.01). The multi-scale entropy MSE1, MSE2, MSE3, MSE4 and MSE5 decreased significantly(P<0.05). When the time-domain features and multi-scale entropy of pulse recordings were used in the modeling, the model has the highest performance, the highest recognition rate(95.67%), the highest specificity(93.99%)and the lowest misjudgment rate(6.01%). Conclusion: To a certain extent, the characteristics of pulse recordings can reflect the pathological state and vascular function of patients with suspected CHD. The screening model for high-risk population of CHD based on the characteristics has potential reference significance.更多还原