随机森林算法在膝骨关节炎穴位敏化研究中的应用初探

Preliminary study on the application of random forest in acupoint sensitization of knee osteoarthritis

目的:采用随机森林算法构建膝骨关节炎(KOA)患者犊鼻穴穴位敏化判别模型,以期为穴位是否敏化的判别及病情评估提供参考。方法:以916例KOA患者和222名健康受试者为研究对象,采集犊鼻穴体表温度、机械痛阈以及压痛阈的数据,运用SPSS 24.0软件进行数据分析,采用Python软件随机森林算法对犊鼻穴穴位敏化数据构建穴位敏化判别模型。结果:与健康受试者比较,KOA患者犊鼻穴温度显著升高、机械痛阈和压痛阈显著降低(P<0.01);穴位敏化判别模型能很好区分KOA患者与健康人(准确率77.85%),其中影像学3级(准确率82.76%)和临床3期(准确率88.37%)KOA患者区分效果最好。结论:穴位敏化判别模型可为穴位是否发生敏化的判别以及病情评估提供参考。

Objective:To construct a sensitization discriminant model of Dubi(ST 35)acupoints in patients with knee osteoarthritis(KOA)by using random deep forest algorithm,in order to provide reference for the judgment of acupoint sensitization and disease assessment.Methods:Taking 916 KOA patients and 222 healthy subjects as the research objects,the data of body surface temperature,mechanical pain threshold and tenderness threshold of Dubi(ST 35)acupoint were collected.SPSS 24.0 software was used for data analysis.The random forest algorithm of Python software was used to construct the acupoint sensitization discriminant model on the sensitization data of Dubi acupoint.Results:Compared with healthy subjects,the temperature of Dubi(ST 35)acupoints in KOA patients increased,and the threshold of mechanical pain and tenderness decreased(P<0.01).The acupoint-sensitized discriminant model could well distinguish KOA patients from healthy people(accuracy rate 77.85%),including imaging grade 3(accuracy rate 82.76%)and clinical stage 3(accuracy rate 88.37%)KOA patients had the best distinction.Conclusion:The acupoint sensitization discriminant model can provide a reference for the judgment of acupoint sensitization and disease assessment.