光谱化学计量分析方法及其在舌诊的应用

发布时间：2018-08-22 12:26

【摘要】：近红外光谱是一种弱吸收谱,在测量分析过程中容易受到诸多因素的影响,不仅包括外界因素,而且包括与样本自身性质特征密切相关的干扰因素。此外,还需要克服由于人体的复杂多变性和个体差异等诸多不确定因素带来的影响。如何处理和利用这些因素,不仅是建立稳健的校正模型以及提高模型预测能力的关键,而且是近红外光谱分析技术进入临床应用的关键。本论文探究了近红外光谱分析过程中各类因素的影响,并系统地将其加以分类,给出了一个应对各类因素的策略,即“M+N”理论。在该理论的指导下,论文对可能会遇到的几种因素(非目标组分、光源电压、温度等)进行了分析与讨论,借助这些因素提出了多种提高校正模型预测能力的方法,并应用到光谱法舌诊的临床实验中。(一)提出了光源电压因素用于提高校正模型预测能力的方法。通过实测变电压光源光谱数据与多组分线性吸收谱仿真模型相结合的方法,验证了该方法用于提高校正模型预测能力可行性。(二)提出了基于温度与光谱关系的定量分析方法。实验结果表明了温度与光谱间关系的存在,并且验证了通过温度与光谱关系的差异可以对样本组分进行定量分析,为研究温度对近红外光谱的影响提供一个新方法。(三)提出了一个校正样本选择的方法,在选择样本时不仅考虑光谱空间、目标变量空间的影响,而且考虑外界因素和内在特征空间的影响。通过两组经典的光谱数据集,从外界因素变量和内在特征变量两个角度验证了所提出方法的可行性与优越性。(四)首次将光谱法舌诊应用到血清胆红素的无创检测中,并取得了较好临床效果。实验结果验证了该方法的可行性,并为其它血液成分的无创检测开辟了新路径。(五)提出了WP-mUVE-LSSVM方法,该方法既可以同时扣除光谱分析中背景与噪声的干扰,又具有非线性功能可以建立稳健的校正模型。首次将该方法应用到糖尿病的快速筛查以及血糖的无创检测中,取得较好的临床效果。论文从基础理论的研究方法着手,以化学计量学方法为依据,分别从外界干扰因素与样本内部特征因素两个方面提出了多种提高校正模型预测能力的方法。并将上述方法应用到光谱法舌诊临床实验中,为光谱法舌诊真正进入临床应用提供了基础,同时对近红外光谱分析技术中其他因素的影响提供了参考与借鉴。
[Abstract]:Near-infrared spectroscopy (NIR) is a kind of weak absorption spectrum, which is easy to be affected by many factors in the process of measurement and analysis, including not only external factors, but also interference factors closely related to the characteristics of the sample itself. In addition, it is necessary to overcome the influence of many uncertain factors, such as complex variability and individual differences. How to deal with and utilize these factors is not only the key to establish robust correction model and improve the prediction ability of the model, but also the key to the clinical application of near-infrared spectroscopy analysis technology. In this paper, the influence of various factors in the process of NIR analysis is explored, and the factors are systematically classified, and a strategy to deal with these factors, namely "M N" theory, is given. Under the guidance of this theory, several factors (non-target components, light source voltage, temperature, etc.) that may be encountered are analyzed and discussed in this paper. With the help of these factors, a variety of methods to improve the prediction ability of the calibration model are proposed. It is applied to the clinical experiment of tongue diagnosis by spectral method. (1) the method of improving the prediction ability of calibration model by using the voltage factor of light source is proposed. The feasibility of using this method to improve the prediction ability of calibration model is verified by combining the measured spectral data of variable-voltage light source with the simulation model of multicomponent linear absorption spectrum. (2) A quantitative analysis method based on the relationship between temperature and spectrum is proposed. The experimental results show the existence of the relationship between temperature and spectrum, and verify that the quantitative analysis of sample components can be carried out by the difference between temperature and spectrum, which provides a new method for studying the influence of temperature on near infrared spectroscopy. (3) A method of correcting sample selection is proposed, which considers not only the influence of spectral space and target variable space, but also the influence of external factors and intrinsic feature space. Through two sets of classical spectral data sets, the feasibility and superiority of the proposed method are verified from the external factor variables and intrinsic characteristic variables. (4) Spectral tongue diagnosis was applied to the noninvasive detection of serum bilirubin for the first time, and good clinical results were obtained. The experimental results verify the feasibility of the method and provide a new path for non-invasive detection of other blood components. (5) WP-mUVE-LSSVM method is proposed, which can not only deduct the interference of background and noise in spectral analysis simultaneously, but also have nonlinear function to establish robust correction model. For the first time, this method has been applied to the rapid screening of diabetes mellitus and noninvasive detection of blood glucose. Based on the method of chemometrics, this paper puts forward several methods to improve the prediction ability of calibration model from two aspects: external interference factors and internal characteristic factors of samples. The above method is applied to the clinical experiment of tongue diagnosis with spectral method, which provides a foundation for the actual application of spectral tongue diagnosis, and also provides a reference for the influence of other factors in the near infrared spectrum analysis technology.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：O657.3;R241

【相似文献】