【作者】 王国强；

【导师】 王与烨；

【作者基本信息】 天津大学 ， 光电子与光子学技术， 2020， 硕士

【摘要】 太赫兹波是频率位于0.1-10 THz内的电磁波,具有低光子能量和指纹谱等特性,在生物医学诊断和物质鉴定等方面应用广泛。衰减全反射模式无需样品预处理,可以对高吸收的样品进行快速检测,得到了科学界的广泛关注。本文针对脑创伤诊断方面的需求和难点,基于太赫兹衰减全反射技术在脑创伤特征性物质表征以及光谱信息处理等方面展开研究工作,主要内容和创新点如下:1.概述了太赫兹衰减全反射光谱技术的发展历程和发展趋势、脑创伤诊断方面的研究现状和挑战,以及太赫兹技术在脑创伤检测中的技术优势。2.研究了太赫兹衰减全反射光谱技术的理论模型。基于衰减全反射原理设计并搭建了太赫兹衰减全反射光谱测量系统;进一步地,从古斯汉森位移、随机误差和系统误差三个方面分析了该系统的误差来源,提出了相应的控制方法。3.基于太赫兹衰减全反射光谱技术对脑创伤相关的特征性物质进行无标记表征研究。利用压缩气体驱动的冲击波产生装置建立了不同程度的冲击性脑创伤大鼠模型,并根据生物医学方法进行了可靠性验证。从组织、体液、总蛋白、生物标志物等不同层次研究脑创伤特征性物质在太赫兹波段的介电响应特性,明确了创伤后24小时内脑脊液、血清以及海马和下丘脑总蛋白的太赫兹波谱特征随时间的变化规律。4.基于多种数据分析方法开展了脑创伤自动识别研究。基于太赫兹相位偏移光谱有效消除了因光纤热漂移和测量随机误差所造成的异常数据,采用特征提取和机器学习中的分类算法对脑脊液和血清的太赫兹光谱数据库进行分析,基于脑脊液和血清的冲击性脑创伤分类准确率最高分别达到了94.4%和95.2%。结果表明,基于太赫兹衰减全反射技术和光谱信息处理技术可以实现灵敏、快速的脑创伤自动诊断。更多还原

【Abstract】 Terahertz(THz)waves refer to electromagnetic waves with frequencies located in the range of 0.1-10 THz,which have characteristics such as low photon energy and fingerprinting.THz technology has been widely used in researches of biomedical diagnostics and substance identification.The attenuated total reflection(ATR)mode requires no sample pretreatment and allows rapid detection of highly absorbed samples.Therefore,it has gained wide attention from the scientific community.To confront the needs and challenges in the diagnosis of traumatic brain injury(TBI),this paper investigated the theoretical models of the THz-ATR technique and blast wave injury mechanism,the characterization of traumatic brain substances,and spectral information processing methods.The main contents and innovations are as follows:1.The history and trends of the development of the THz-ATR technique,the current status and challenges of research in diagnosis of TBI,and the technical advantages of the THz technique in TBI detection were summarized and outlined.2.The theoretical models of the THz-ATR were investigated.Based on the principle of total internal reflection,the THz-ATR spectroscopy measurement system was designed and built.Important parameters in the THz spectroscopy were deduced.Further,the sources of measurement errors are discussed in terms of Goos–H?nchen displacement,random error and systematic error.Corresponding error-control methods were also proposed.3.Based on THz-ATR spectroscopy,the label-free characterization study was performed on the characteristic substances associated with blast-induced TBI(b TBI).Based on a compressed-gas driven blast wave generator,different levels of b TBI model was developed in rat.Its reliability was verified via biomedical methods.The dielectric response characteristics of b TBI substances in the terahertz band were studied at different levels,including tissues,fluid,total protein,and biomarkers.The timedependent changes of the THz spectral characteristics of cerebrospinal fluid,serum,hippocampal and hypothalamic total protein were clarified in the 24 hours after blast exposure.4.An automatic TBI identification study was carried out based on multiple spectral data analysis methods.Outliers caused by fiber thermal drift and measurement random errors were effectively eliminated based on statistical distribution of phase-shift spectra.The terahertz spectral database of cerebrospinal fluid and serum was analyzed using feature extraction and classification algorithms in machine learning.The developed classification models have good generalization performance.The highest accuracy rates of 94.4% and 95.2% were achieved for CSF-based and serum-based classification of b TBI,respectively.The results suggested that sensitive and rapid automatic diagnosis of brain trauma can be achieved based on THz-ATR and spectral information processing techniques.更多还原