摘要
目的舌诊是2型糖尿病辨证的重要方法之一,根据糖尿病足患者舌象提出一种糖尿病足Wagner分级的图像识别方法。方法通过标准化舌象采集获得舌图,之后提取舌尖、舌边、舌中、舌根部舌质、舌苔区域的红(Red)、绿(Green)、蓝(Blue)颜色空间值,最后使用极端梯度提升(XGBoost)、随机森林(Random Forest,RF)、支持向量机(Support Vector Machine,SVM)算法作为分类器训练识别数据,并横向进行对比。结果通过对88例数据训练识别,在数据预处理基础上按照7∶3划分训练集和测试集,得出随机森林与XGBoost的方法远高于支持向量机,在舌尖部两者表现最优,舌两边随机森林更好,但在舌根部XGBoost方法在正确率、精确度、F1值方面均高于随机森林。结论在糖尿病足Wagner分级舌图识别方面,XGBoost方法是一种值得推广与应用的图像识别方法。
Objective Tongue diagnosis is one of the important methods for the differentiation of type 2 diabetes,and an image recognition method for Wagner grading of diabetic foot is proposed according to the tongue image of diabetic foot patients.Methods Obtain tongue maps by standardizing tongue image collection,and then the RGB value of tongue tip,tongue edge,tongue,tongue root tongue,tongue moss area is extracted,and finally the support vector machine,random forest(RF),extreme gradient enhancement XGBoost algorithm is used as classifier training identification data,and the horizontal comparison is made.Results Through the identification of 88 cases of data training,the training set and test set are divided according to 7∶3 on the basis of data pre-processing,and the method of random forest and XGBoost is much higher than that of support vector machine,which is the best performance in the tip of the tongue,and the random forest on both sides of the tongue is better,but the XGBoost method at the root of the tongue is higher than the random forest method in terms of accuracy,accuracy and F1 value.Conclusion The XGBoost method is an image recognition method worthy of promotion and application in the recognition of diabetic foot Wagner graded tongue chart.
作者
田之魁
王东军
李生启
关媛媛
孙璇
朱青青
王泓午
Tian Zhikui;Wang Dongjun;Li Shengqi;Guan Yuanyuan;Sun Xuan;Zhu Qingqing;Wang Hongwu(College of Health Science and Engineering,Tianjin University of Traditional Chinese Medicine,Tianjin 301617,China;School of Computing,Nankai University,Tianjin 301617,China)
出处
《世界科学技术-中医药现代化》
CSCD
北大核心
2023年第4期1442-1446,共5页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
国家科学技术部国家重点研发计划(2017YFC1703305):中医舌象诊断规范与标准研究及应用
负责人:王泓午
作者简介
通讯作者:王泓午,教授,主要研究方向:中医证候标准化及客观。
