用于反射式荧光成像的光谱分离方法
发布时间：2018-09-14 14:24
【摘要】：反射式荧光成像技术可以连续、无创、高通量地在体检测小动物体内被标记的细胞和分子，追踪各种疾病的形成和发展。然而这种成像方式的缺陷在于在体成像时皮肤和食物自发荧光的存在会大大降低系统的探测灵敏度，使感兴趣荧光团难以准确监测和定位。另外，为了同时监测多种生物过程，，需要利用多种荧光标记物标记不同的分子进行荧光成像。这些荧光团光谱混叠，无法独立分辨它们各自的信息。而多光谱分离法可用于反射式荧光成像时自发荧光的去除和多种感兴趣荧光团的分离。 本文提出一种多光谱分离方法：从5-6幅多光谱荧光图像中在体提取自发荧光和感兴趣荧光团的纯光谱数据后，再使用线性分离算法去除自发荧光，并区分不同的目标荧光团。将本算法运用到反射式荧光成像系统中，去除了自发荧光的影响，实现了分别表达TagRFP和mLumin荧光蛋白的两种BL21大肠杆菌样品的分离。这两种荧光团与自发荧光的信噪比在分离前后分别从9.23dB和4.70dB提高到35.69dB和24.91dB。此外，在感兴趣荧光团的信号较微弱以致于其空间分布无法预测的情况下，本文对上述算法进行了改进。首先用初始化中心点的分类算法对原始多光谱荧光图像按光谱性质的不同进行分类，获得感兴趣荧光团以及自发荧光的空间分布后，再提取各个荧光团的纯光谱用于线性分离算法。通过在体模型实验和在体鼻咽癌肿瘤模型实验进一步验证了改进后线性分离算法的可行性。
[Abstract]:The reflective fluorescence imaging technique can continuously, noninvasively and high-throughput detect labeled cells and molecules in small animals and track the formation and development of various diseases. However, the defect of this imaging method is that the presence of skin and food autofluorescence in volume imaging will greatly reduce the detection sensitivity of the system and make it difficult for interested fluorescence groups to accurately monitor and locate. In addition, in order to monitor multiple biological processes simultaneously, a variety of fluorescent markers are used to label different molecules for fluorescence imaging. The spectra of these fluorescence clusters are overlapped and their respective information cannot be identified independently. The multi-spectral separation method can be used for the removal of autofluorescence and the separation of a variety of interesting fluorescence groups in reflective fluorescence imaging. In this paper, a multispectral separation method is proposed: after extracting in vivo the pure spectral data of autofluorescence and interesting fluorescence groups from 5-6 multispectral fluorescence images, the linear separation algorithm is used to remove the autofluorescence and distinguish different target fluorescence groups. The algorithm is applied to the reflective fluorescence imaging system to remove the influence of autofluorescence and to separate two kinds of BL21 Escherichia coli samples expressing TagRFP and mLumin fluorescent proteins respectively. The signal-to-noise ratio of these two groups increased from 9.23dB and 4.70dB to 35.69dB and 24.91 dB before and after separation. In addition, under the condition that the signal of the fluorescence group of interest is weak and its spatial distribution can not be predicted, the above algorithm is improved in this paper. First, the original multispectral fluorescence images are classified according to the spectral properties using the classification algorithm of initializing the center points, and the spatial distribution of the interesting fluorescence groups and the autofluorescence are obtained. Then the pure spectrum of each fluorescence group was extracted for linear separation algorithm. The feasibility of the improved linear separation algorithm is further verified by in vivo model experiments and in vivo tumor model experiments for nasopharyngeal carcinoma (NPC).
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R310
本文编号：2242965
[Abstract]:The reflective fluorescence imaging technique can continuously, noninvasively and high-throughput detect labeled cells and molecules in small animals and track the formation and development of various diseases. However, the defect of this imaging method is that the presence of skin and food autofluorescence in volume imaging will greatly reduce the detection sensitivity of the system and make it difficult for interested fluorescence groups to accurately monitor and locate. In addition, in order to monitor multiple biological processes simultaneously, a variety of fluorescent markers are used to label different molecules for fluorescence imaging. The spectra of these fluorescence clusters are overlapped and their respective information cannot be identified independently. The multi-spectral separation method can be used for the removal of autofluorescence and the separation of a variety of interesting fluorescence groups in reflective fluorescence imaging. In this paper, a multispectral separation method is proposed: after extracting in vivo the pure spectral data of autofluorescence and interesting fluorescence groups from 5-6 multispectral fluorescence images, the linear separation algorithm is used to remove the autofluorescence and distinguish different target fluorescence groups. The algorithm is applied to the reflective fluorescence imaging system to remove the influence of autofluorescence and to separate two kinds of BL21 Escherichia coli samples expressing TagRFP and mLumin fluorescent proteins respectively. The signal-to-noise ratio of these two groups increased from 9.23dB and 4.70dB to 35.69dB and 24.91 dB before and after separation. In addition, under the condition that the signal of the fluorescence group of interest is weak and its spatial distribution can not be predicted, the above algorithm is improved in this paper. First, the original multispectral fluorescence images are classified according to the spectral properties using the classification algorithm of initializing the center points, and the spatial distribution of the interesting fluorescence groups and the autofluorescence are obtained. Then the pure spectrum of each fluorescence group was extracted for linear separation algorithm. The feasibility of the improved linear separation algorithm is further verified by in vivo model experiments and in vivo tumor model experiments for nasopharyngeal carcinoma (NPC).
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R310
【参考文献】
相关期刊论文 前5条
1 王荣福;肿瘤核素显像的临床应用研究[J];北京医学;2004年05期
2 杨阔;张小琴;宋永;秦天莺;;分子成像技术及应用[J];河南教育学院学报(自然科学版);2010年04期
3 宫彦军,王艳红,禹秉熙;高光谱识别目标的光谱分离分析方法[J];内蒙古大学学报(自然科学版);2003年02期
4 庄天戈;;走近分子成像[J];中国医疗器械杂志;2007年02期
5 种敏琪;秦斌杰;;生物荧光谱分离端元提取算法的实现与比较[J];中国医疗器械杂志;2010年04期
本文编号：2242965
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2242965.html
