摘要
目的提出基于舌象和形体特征的中医体质辨识模型,探讨中医以舌辨质客观化、规范化方法。方法提取客观化的舌象特征(舌质与舌苔的色度和饱和度、舌苔纹理的平均亮度、平滑度以及齿痕特征)和形体特征,建立基于神经网络和支持向量机的客观化舌象和形体特征的辅助中医体质识别模型。结果对平和质、气虚质、阴虚质和气淤质四种体质进行模型的训练和测试,结果表明融合舌象特征和形体特征的中医体质辨识模型能有效地辅助中医体质识别,且支持向量机对四种体质辨识效果总体上优于神经网络。结论基于客观化的舌象特征和形体特征辅助中医体质辨识有利于提高中医体质辨识的客观化水平,合理选择机器学习算法可以提高中医体质辨识的准确性。
Objective A Traditional Chinese Medical(TCM)physique identification models based on tongue and body features has been propose in order to improve the objectification and standardization method of TCM physique identification.Methods The characteristics of objective tongue image(the features of hue and saturation of tongue body and tongue coating,average intensity,smoothness of tongue coating texture and tongue marks)and body feature were extracted and analyzed based on the auxiliary TCM physique identification model of artificial neural network and support vector machine.Results The training and testing of the four physiques of pinghe、qixu、yinxu and qiyu shows that the TCM physique identification model with the characteristics of the tongue and body features can effectively assist the TCM physique recognition.Comparing with artificial the neural network,the support vector machine show a better effects of identification for the four constitution.Conclusion It is beneficial to improve the objectification level of TCM constitution identification based on the characteristics of the tongue and body features.And appropriate machine learning algorithm would improve the accuracy of TCM constitution identification.
作者
潘思行
林育
周苏娟
黄展鹏
Pan Sixing;Lin Yu;Zhou Sujuan;Huang Zhanpeng(School of Public Health,Guangdong Pharmaceutical University,Guangzhou,510006,China;Institute of Clinical Pharmacology,Guangzhou University of Chinese Medicine,Guangzhou,510006,China;School of Medical Information Engineering,Guangdong Pharmaceutical University,Guangzhou,510006,China)
出处
《世界科学技术-中医药现代化》
CSCD
北大核心
2020年第4期1341-1347,共7页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
广东省中医药局建设中医药强省基金面上项目(20151266):融合望诊图像特征的高脂血症神经网络预警模型研究,负责人:周苏娟
关键词
舌象特征
形体特征
神经网络
支持向量机
Tongue feature
Body feature
Artificial Neural Network
support Vector Machine
作者简介
通讯作者:黄展鹏,硕士研究生导师,副教授,主要研究方向:图像处理,中医诊断客观化研究.
