基于电子鼻的2型糖尿病常见病位的气味图谱辨识研究  

Research on recognition of odor response patterns of common disease locations of type 2 diabetes mellitus based on electronic nose

作　　者：林雪娟[1,2] 周福 吴青海[1,2] 田群辉 罗志明 王永发 李灿东[1,2] LIN Xue-juan;ZHOU Fu;WU Qing-hai;TIAN Quan-hui;LUO Zhi-ming;WANG Yong-fa;LI Can-dong(Fujian University of Traditional Chinese Medicine,Fuzhou 350122,China;Key Laboratory of Chinese Medicine Health Status Differentiation of Fujian Province(Fujian University of Traditional Chinese Medicine),Fuzhou 350122,China;Xiamen University,Xiamen 361005,China;Jinjiang Hospital of Traditional Chinese Medicine Affiliated to Fujian University of Traditional Chinese Medicine,Jinjiang 362201,China)

机构地区：[1]福建中医药大学,福州350122 [2]福建省中医健康状态辨识重点实验室(福建中医药大学),福州350122 [3]厦门大学,厦门361005 [4]福建中医药大学附属晋江中医院,晋江3622011

出　　处：《中华中医药杂志》2022年第7期3785-3789,共5页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家自然科学基金项目(No.81973752);福建省自然科学基金项目(No.2018J01892);载人航天领域预先研究项目(No.020104)。

摘　　要：目的:运用电子鼻技术探讨2型糖尿病前期与糖尿病期患者及其常见病位的口腔呼气的气味图谱辨识。方法:选择61例2型糖尿病前期患者、165例2型糖尿病期患者和50例健康者,运用基于阵列式气体传感器技术的电子鼻(EN011103-A)采集口腔呼气的气味图谱,采用模式识别的方法进行气味图谱辨识。结果:①2型糖尿病前期常见病位证素分布由高到低依次为肝、肾、肺、脾;糖尿病期常见病位证素分布由高到低依次为肾、肝、脾、胃、肺、心、心神。②使用深度神经网络算法对糖尿病前期、糖尿病期和健康组的气味图谱辨识的平均分类准确率最高,为98.19%。③使用随机森林算法对糖尿病前期、糖尿病期常见病位证素的气味图谱辨识的平均分类准确率均是最高,分别为76.00%、80.76%。④除病位证素心外,使用模式识别方法对单个病位证素气味图谱的辨识准确率均达到80%以上。结论:运用阵列式气体传感器电子鼻结合模式识别算法可以对2型糖尿病的分期与病位作初步的辨识,可以为中医嗅诊客观化研究提供新的手段,为中医病证诊断提供新的方法。Objective: To explore the recognition of oral breath odor response patterns of prediabetic and diabetic patients of type 2 diabetes mellitus with their common disease locations by electronic nose. Methods: A total 61 prediabetic patients of type 2 diabetes mellitus and 165 type 2 diabetic patients and 50 healthy persons were observed. Odor response patterns of oral breath were collected by the electronic nose(EN0l1103-A) based on a film of gas sensor array. The method of pattern recognition was used to recognize the odor response pattern. Results:①The distribution of common syndrome elements of disease location of prediabetic patients of type 2 diabetes mellitus from high to low was liver, kidney, lung, spleen. The distribution of common syndrome elements of disease location of diabetic patients from high to low was kidney, liver, spleen, stomach,lung, heart and mind.②Odor response patterns of prediabetic and diabetic patients and healthy persons were identified by the method of pattern recognition. The average classification accuracy by deep neural networks was the highest, which was 98.19%.③The average classification accuracy of odor response patterns of common disease locations of prediabetic patients or diabetic patients by random forest were the highest, which were 76.00%, 80.76%, respectively.④Most of the identification accuracy of odor response patterns of syndrome elements of disease location was higher than 80% except heart. Conclusion: The stage and location of type 2 diabetes mellitus could be recognized preliminary by the electronic nose based on a film of gas sensor array. The electronic nose could provide a kind of new means for the study on the objectification of TCM smelling examination and provide a method for the diagnosis of TCM disease and syndrome.

关 键 词：嗅诊 电子鼻 2型糖尿病 病位 气味图谱 

分 类 号：R259[医药卫生—中西医结合]