中华中医药学刊2024,Vol.42Issue(5) :17-19.DOI:10.13193/j.issn.1673-7717.2024.05.004

基于深度卷积神经网络算法和先验知识构建冠心病患者大鱼际望诊模型的思路与方法

Thoughts and Methods about Deep Learning Model of Thenar Inspection for Coronary Heart Disease Patients Based on Deep Convolutional Neural Network and Prior Knowledge

刘大胜 李玉坤 赵志伟 孙晨格 杨伟 王丽颖 韩学杰
中华中医药学刊2024,Vol.42Issue(5) :17-19.DOI:10.13193/j.issn.1673-7717.2024.05.004

基于深度卷积神经网络算法和先验知识构建冠心病患者大鱼际望诊模型的思路与方法

Thoughts and Methods about Deep Learning Model of Thenar Inspection for Coronary Heart Disease Patients Based on Deep Convolutional Neural Network and Prior Knowledge

刘大胜 1李玉坤 1赵志伟 1孙晨格 2杨伟 1王丽颖 1韩学杰1
扫码查看

作者信息

  • 1. 中国中医科学院中医临床基础医学研究所,北京 100700
  • 2. 中国中医科学院中医临床基础医学研究所,北京 100700;陕西中医药大学,陕西西安 712046
  • 折叠

摘要

基于全息理论的中医望诊可以辅助诊断西医疾病,但目前中医望诊主要依靠名老中医药专家的经验传承,存在望诊客观化、标准化程度不够,缺乏行业内认可度高的望诊转化技术的问题.而望诊融合人工智能信息化技术,可以提升中医望诊客观化、标准化的水平,可以有效地降低疾病的恶化率和病死率,促进中医望诊经验的转化.据此,结合前期开展的大鱼际特征与冠心病关系研究,得出大鱼际望诊可以用于冠心病早期预警筛查.以大鱼际望诊和冠心病之间的关系为例,将先验知识和深度卷积神经网络算法深度融合,将特征提取和分类合为一体,利用深度学习端对端的显著特点,输入观察到的原始大鱼际图像像素数据或信息,通过对大鱼际照片的大量深度学习,构建冠心病患者的关键特征要素,融合先验知识后,输出是否为冠心病的分类结果,中间为深层的网络结构.这一思路将提出一种中医望诊客观化、标准化的智能化算法,促进中医望诊经验的转化思路与方法,以提高基层群众的疾病预警筛查能力,服务"健康中国"战略.

Abstract

Inspection of traditional Chinese medicine(TCM)based on the holographic theory can assist in the diagnosis of Western diseases.However,the current TCM inspection mainly relies on the experience of famous and old experts,and there is the problem of insufficient objectification and standardization of inspection,and the lack of a highly recognized inspection transfor-mation technology in the industry.The integration of inspection with artificial intelligence information technology can enhance the objectivity and standardization of TCM inspection,which can effectively reduce the deterioration rate and mortality of diseases and promote the transformation of TCM inspection experience.Taking the relationship between large fissure lookout and coronary heart disease(CHD)as an example,the prior knowledge and deep convolutional neural network algorithm are deeply fused to combine feature extraction and classification into one,using the significant features of deep learning end-to-end,inputting the observed pixel data or information of the original large fissure image,constructing the key feature elements of CHD patients through exten-sive deep learning of large fissure photos,and outputting after fusing the prior knowledge whether it is the classification result of CHD,and the middle is the deep network structure.This idea will propose an intelligent algorithm for the objectification and standardization of TCM inspection,promote the transformation of TCM inspection,improve grassroots people s ability of disease early warning and screening,and serve the strategy of"Healthy China".

关键词

图像信息/深度卷积神经网络/先验知识/大鱼际望诊/冠心病

Key words

image identification/deep convolutional neural network/prior knowledge/thenar inspection/coronary heart dis-ease

引用本文复制引用

基金项目

国家自然科学基金青年科学基金(82205320)

中国中医科学院优秀青年科技人才培养专项(ZZ16-YQ-035)

出版年

2024
中华中医药学刊
中华中医药学会 ,辽宁中医药大学

中华中医药学刊

CSTPCDCSCD北大核心
影响因子:1.007
ISSN:1673-7717
参考文献量38
段落导航相关论文