基于呼出气体检测口臭等级的电子鼻研究  

Research on Electronic Nose Based on Exhaled Gas to Detect Halitosis Grade

作　　者：张涛 薛莹莹 陈远涛 熊忆舟 于伟杰 张钧煜 万浩 张若彤 叶玮[2] 王平[1] ZHANG Tao;XUE Yingying;CHEN Yuantao;XIONG Yizhou;YU Weijie;ZHANG Junyu;WAN Hao;ZHANG Ruotong;YE Wei;WANG Ping(Biosensor National Special Laboratory,Key Laboratory for Biomedical Engineering of Education Ministry,Department of Biomedical Engineering,Zhejiang University,Hangzhou Zhejiang 310027,China;Shanghai Ninth People’s Hospital,School of Medicine,Shanghai Jiao Tong University,Shanghai 200011,China)

机构地区：[1]浙江大学生物传感器国家专业实验室,生物医学工程教育部重点实验室,生仪学院,浙江杭州310027 [2]上海交通大学医学院附属第九人民医院,上海200011

出　　处：《传感技术学报》2021年第7期867-873,共7页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(31627801,61901412)。

摘　　要：口臭对人们的社交互动和生活质量产生严重的负面影响,当前对口臭的诊断方法难以在个人及家庭中应用。本文通过采集口臭病人样本,建立了基于GC-MS技术检测口气中VSCs的标准化流程,采用ROC曲线分析方法确定了不同感官评分的VSCs浓度范围。在此基础上,研制了一套基于电化学传感器并结合卷积神经网络模型的电子鼻,用于口臭标志物检测和口臭等级的判别。该电子鼻对硫化氢和甲硫醇的检出限分别为39×10^(-9)和48×10^(-9)。通过与GC-MS检测结果比较,该电子鼻检测实际临床样本中VSCs的平均相对误差为11.4%,识别口臭等级的正确率达到80%,表明该电子鼻可以满足口臭分级检测的临床要求,具有良好的应用前景。Halitosis can have a serious negative impact on people’s social interaction and quality of life and the current diagnosis methods of halitosis are difficult to be applied in individuals and families. By collecting samples from patients with halitosis, a standardized process for detecting volatile sulfur compounds in breath based on gas chromatography and mass spectrometry was established, and the receiver operating characteristic curve was used to determine the VSCs concentration range for different organoleptic scores. On this basis, a set of electronic noses based on electrochemical sensors combined with a convolutional neural network model were developed for the detection of halitosis markers and the judgment of the level of halitosis. The detection limits of the electronic nose for hydrogen sulfide and methyl mercaptan are 39×10^(-9) and 48×10^(-9),respectively. By comparing with the GC-MS measurement results, the average relative error of VSCs in the actual clinical samples detected by the electronic nose is 11.4%,and the correct rate of identifying the halitosis level is 80%,indicating that the electronic nose can meet the clinical requirements of halitosis grading detection.

关 键 词：电子鼻 口臭 呼出气体 挥发性含硫化合物 卷积神经网络 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] R781[自动化与计算机技术—控制科学与工程]