【作者】 张涛；

【导师】 王平；

【作者基本信息】 浙江大学 ， 工程硕士（专业学位）， 2021， 硕士

【摘要】 口臭是指呼吸时从口腔中散发出的令人讨厌的气味,会对个人形象及身心健康带来严重的影响。口臭的出现通常与口腔卫生状况和口源性疾病相关,当前已经成为口腔门诊中除龋病和牙周疾病以外主诉最多的疾病。商用的口臭检测仪器由于成本、操作复杂程度和体积等因素使用范围只能局限于医院中,因此设计一种具有个人或家庭使用前景的口臭标志物检测及口源性疾病筛查的电子鼻具有重要的意义。基于这个出发点,本文工作的主要内容和创新点如下:1、建立了口气样本采集及气相色谱与质谱联用技术(Gas Chromatography-Mass Spectroscopy,GC-MS)检测口腔内挥发性硫化物(Volatile Sulfur Compounds,VSCs)的标准化方法和流程。硫化氢和甲硫醇作为口源性口臭的潜在标志物已经被逐渐地应用于临床上对口臭的评估,但是由于它们的高反应性和痕量浓度导致对呼出气体中的VSCs进行分析时具有挑战性。针对这些难点,本文利用化学性能稳定的Tedlar采样袋收集口气样本,并基于GC-MS技术,采用单离子检测扫描模式及外标法对呼出气体中的VSCs进行了定性和定量分析。2、对Halimeter、Oral Chroma和GC-MS三种口臭检测仪器的测量结果进行了分析,确定了不同感官评分间的VSCs浓度范围并建立了口源性疾病诊断模型。每个入选的临床对象都经过了Halimeter、Oral Chroma和GC-MS三种口臭诊断仪器的检测。本文分析了不同仪器检测结果之间的相关性以及检测结果与感官评分结果之间的相关性,并采用受试者工作特征曲线(Receiver Operating Characteristic Curve,简称ROC曲线)的分析方法确定了不同感官评分的VSCs浓度范围。以GC-MS的测量结果作为疾病诊断模型的自变量,设计了基于线性判别分析和逻辑回归方法的口源性疾病诊断模型,并对不同模型的诊断效果进行了分析和比较。3、以GC-MS的测量结果作为传感器筛选的参考标准,完成了基于呼出气体进行口臭标志物检测及口源性疾病筛查的电子鼻的软硬件设计。从GC-MS测量的结果中确定了口臭患者呼出气体中VSCs的浓度范围,以GC-MS的测量结果作为电子鼻传感器筛选的参考标准,筛选了量程、灵敏度、分辨率和特异性符合检测要求的电化学传感器作为电子鼻的核心检测元件,在此基础上设计了传感器阵列的气室及气路结构。开发了基于呼出气体进行口臭标志物检测及疾病筛查的电子鼻系统,详细分析了电路设计中硬件参数对电子鼻性能的影响,并完成了电子鼻下位机程序和上位机数据采集软件的设计。4、对电子鼻的基本性能进行了测试,构建了基于一维卷积神经网络的VSCs定量算法模型,并评估了电子鼻在实际临床样本检测过程中的效果。使用配气系统配制作不同浓度的标准气体对电子鼻的检出限、重复性和线性度进行了测试,结果显示电子鼻对硫化氢和甲硫醇的检出限分别为39ppb和48ppb,并具有优异的重复性和线性度。构建了基于一维卷积神经网络的VSCs定量算法模型。采集了27个临床样本同时使用电子鼻和GC-MS检测,与GC-MS检测结果相比,该电子鼻检测实际临床样本中VSCs的平均相对误差为11.7%。以辨嗅员的感官评分作为金标准,将电子鼻的评分与辨嗅员的感官评分结果进行对比,该电子鼻评分的正确率为77.8%。将该电子鼻的测量结果作为自变量代入基于线性判别分析原理的口源性疾病诊断模型,在不考虑厚舌苔因素的情况下,该电子鼻区分口源性疾病患者与健康人的灵敏度和特异度分别为87.5%和72.7%,整体正确率为81.5%。更多还原

【Abstract】 Halitosis refers to the unpleasant smell emitted from the mouth when breathing,which will have a serious impact on personal image and mental health.The appearance of halitosis is usually related to oral hygiene and oral diseases.Nowadays,halitosis has become the most common complaint disease except caries and periodontal diseases.Because of the cost,operation complexity and volume of commercial halitosis detection instruments,their application scope can only be limited to hospitals.Therefore,it is of great significance to design an electronic nose for halitosis marker detection and oral disease screening with personal or family use prospects.Based on this starting point,the main contents and innovations of this paper are as follows:1.A standardized method and process for the determination of volatile sulfide compounds in oral cavity by gas chromatography-mass spectrometry(GC-MS)was established.Hydrogen sulfide and methyl mercaptan as potential markers of halitosis have been gradually used in clinical evaluation of halitosis.However,due to their high reactivity and trace concentration,the analysis of VSCs in exhaled gas is challenging.In view of these difficulties,102 breath samples were collected by using Tedlar sampling bag.Based on GC-MS technology,VSCs in exhaled breath were analyzed qualitatively and quantitatively by using single ion monitoring mode and external standard method.2.The measurement results of Halimeter,Oral Chroma and GC-MS were analyzed to determine the concentration range of VSCs between different organoleptic scores and establish the diagnosis model of oral diseases.All subjects were tested by Halimeter,Oral Chroma and GC-MS.This paper analyzes the correlation between the test results of different instruments and the correlation between the test results and organoleptic scores,and uses receiver operating characteristic curve(ROC curve)to determine the concentration range of VSCs with different organoleptic scores.Taking the measurement results of GC-MS as the independent variables of the disease diagnosis model,a diagnosis model of oral diseases based on linear discriminant analysis and logistic regression method was designed,and the diagnosis effects of different models were analyzed and compared.3.Taking the measurement results of GC-MS as the reference standard of sensor screening,the software and hardware design of the electronic nose based on exhaled gas for halitosis marker detection and disease screening was completed.The concentration range of VSCs in the exhaled gas of patients with halitosis was determined from the results of GC-MS measurement.The results of GC-MS were used as the reference standard for the screening of electronic nose sensors.The electrochemical sensors with range,sensitivity,resolution and specificity in accordance with the requirements of the detection were selected as the core detection elements of electronic nose.On this basis,the air chamber and air circuit structure of sensor array are designed.The electronic nose system based on breath gas for odor detection and disease screening is developed.The influence of hardware parameters on electronic nose performance is analyzed in detail.The design of the program of electronic nose slave computer and data acquisition software of host computer are completed.4.The basic performance of the electronic nose is tested,the quantitative algorithm model of VSCs based on one-dimensional convolutional neural network is constructed,and the effect of the electronic nose in the detection of actual clinical samples is evaluated.The detection limit,repeatability and linearity of the electronic nose were tested with different concentrations of standard gases prepared by the gas distribution system.The results show that the detection limit of the electronic nose for hydrogen sulfide and methyl mercaptan are 39 ppb and 48 ppb respectively,and it has excellent repeatability and linearity.A quantitative algorithm model of VSCs based on one-dimensional convolutional neural network was constructed.27 clinical samples were collected and detected by electronic nose and GC-MS simultaneously.Compared with the results of GC-MS,the average relative error of the electronic nose in detecting VSCs in actual clinical samples was 11.7%.Taking the sensory score of olfactory discriminator as the gold standard,the electronic nose score was compared with the sensory score of olfactory discriminator,and the accuracy of the electronic nose score was 77.8%.Taking the measurement results of the electronic nose as independent variables and substituting it into the oral disease diagnosis model based on linear discriminant analysis,the sensitivity and specificity of the electronic nose to distinguish oral disease patients and healthy people were 87.5% and 72.7% respectively,and the overall accuracy rate was 81.5%.更多还原