基于司外揣内原理的中医“舌象-体质”深度学习模型构建研究  

Research on the Construction of Deep Learning Model of“Tongue Manifestation-Physical Constitution”in Traditional Chinese Medicine Based on the Principle of“Governing Exterior to Infer Interior”

作　　者：姜荣荣[1] 高佳奕 杨涛[2,3] JIANG Rong-rong;GAO Jia-yi;YANG Tao(School of Nursing,Nanjing University of Chinese Medicine,Nanjing 210023,Jiangsu,China;School of Artificial Intelligence and Information Technology,Nanjing University of Chinese Medicine,Nanjing 210023,Jiangsu,China;Jiangsu Provincial Engineering Research Center of TCM External Medication Development and Application,Nanjing 210023,Jiangsu,China)

机构地区：[1]南京中医药大学护理学院,江苏南京210023 [2]南京中医药大学人工智能与信息与信息技术学院,江苏南京210023 [3]江苏省中医外用药开发与应用工程研究中心,江苏南京210023

出　　处：《医学信息》2022年第18期10-13,共4页Journal of Medical Information

基　　金：江苏高校哲学社会科学研究项目(编号:2020SJA0320);国家自然科学基金(编号:82174276);中国博士后科学基金会(编号:2021M701674);江苏省博士后科研资助计划项目(编号:2021K457C);江苏高校“青蓝工程”资助(编号:2021);教育部产学合作协同育人项目(编号:202101224003、201902183039);江苏省大学生创新创业训练计划创新项目(编号:202210315103Z)。

摘　　要：司外揣内是中医诊断和健康评估的重要手段,中医体质辨识是这一理论的典型应用,传统的中医体质辨识依赖中医体质评分量表进行判定,效率低下。本文将深度学习技术应用到中医体质辨识领域,基于司外揣内原理,将多尺度舌象信息融入注意力模块中,以残差方式堆叠,将图像模态和文本模型的舌象信息融合,构建中医体质辨识模型,并在数据集上进行训练和测试,以期实现较为精准的体质判别。Governing exterior to infer interior is an important means of traditional Chinese medicine diagnosis and health assessment,traditional Chinese medicineconstitution identification is a typical application of this theory.However traditional Chinese medicine constitution identification relies on traditionalChinese medicine constitution rating scale,which is inefficient.In this paper,deep learning technology is applied to the field of traditional Chinese medicine constitution identification.Based on the principle of inferring the inside from the outside,the multi-scale tongue manifestation information is integrated into the attention module,stacked in a residual manner,and the tongue manifestation information of the image mode and the text model is fused to construct the traditionalChinese medicine constitution identification model.Training and testing are performed on the data set in order to achieve more accurate constitution discrimination.

关 键 词：司外揣内 中医舌象 中医体质 深度学习 体质辨识模型 

分 类 号：R2-03[医药卫生—中医学] TP391.1[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]