脑血管光片显微成像方法与图像处理算法研究

CASIA OpenIR > 毕业生 > 硕士学位论文

	脑血管光片显微成像方法与图像处理算法研究
	梁潇
	2016-05-29
学位类型	工学硕士
中文摘要	光片显微成像是一种高时空分辨率的三维荧光成像方式，具有成像速度快、低光毒性与低光漂白性的优点。由于其能够在不破坏样本完整性的前提下对活体生物如斑马鱼及离体组织等在细胞水平上进行长时间观测，光片显微成像技术在基础生物学尤其是神经科学领域得到广泛应用。光片显微成像介观尺度的分辨率能够观测到包括毛细血管在内的小鼠脑血管网络。探索小鼠脑血管网络的光片显微成像方法，获取高对比度的血管图像，对于研究脑血管相关疾病与生理过程有重要意义，是一个具有重要价值的研究方向。然而局部高散射与高吸收引起的条纹伪影问题，严重降低脑血管光片显微成像图像质量，限制光片显微成像应用范围，有必要研究针对光片显微成像条纹伪影去除算法。小鼠脑血管光片显微成像存在微血管信噪比低的缺点，需要研究血管增强算法，提高微血管对比度。本文以光片显微成像为中心开展了系列研究，将样本透明化染色处理、光片图像处理算法、光片成像生物应用串成一条研究主线。工作主要集中于光片显微成像小鼠脑血管成像方法与应用、光片图像条纹伪影去除与血管增强算法研究研究。本文的主要研究内容包括以下几点： 1. 研究小鼠脑血管光片显微成像方法与生物应用。结合BABB透明化与免疫组化方法设计了iBABB方法。以CD31与PDGFRβ为靶点，对小鼠脑血管内皮细胞与周细胞进行标记，实现利用光片显微成像系统基于血管内皮细胞与周细胞同时对小鼠脑血管网络进行成像，探索脑部血管内皮细胞与周细胞关系。 2. 设计实现了光片显微成像图像条纹伪影除去算法。针对条纹伪影影响成像质量的问题，结合非下采样轮廓波变换与频域滤波，设计条纹伪影去除算法，能够同时去除单方向与多方向条纹伪影，并通过对比试验验证算法，取得了很好的去除条纹伪影效果。 3. 设计实现血管增强算法。针对脑血管成像中微血管信噪比低的问题，引入L1范数设计了基于梯度改进的血管增强算法，显著增强光片图像脑血管信号与对比度。算法依据原图像与其梯度图像，构建L1范数最优化模型，采用迭代Split Bregman方法求解得到增强血管图像。
英文摘要	Light sheet microscopy (LSM) is a high temporal-spatial resolution 3D fluorescence imaging technique, which has the advantages of low phototoxicity and low bleaching. Owing to the ability of long-term imaging of living samples such as zebra fish and in vitro tissue at cellular level without broken them, LSM has been widely used in basic biology area especially in system neurosciences. Moreover LSM’s mesoscale resolution makes it possible to observe mouse brain micro-vessel network. Exploring the method of observing mouse brain vessel network with LSM and acquiring high contrast vascular images has important significance for research of cerebrovascular related diseases and physiological process and thus to be a meaningful research field of great value. However stripes caused by scattering or/and refraction phenomenon in light path drastically deteriorates the image quality limits the application of LSM. Thus there is great demand to develop LSM destripe method. Mouse brain vascular imaging with LSM has the drawback of low signal to noise ratio, so it is need to develop vessel enhance method to improve its contrast. This thesis mainly does research on LSM, including sample clearing and staining, LSM image processing and biological applications. Mouse brain vessel network imaging method and application with LSM has also been studied. Destripe and vascular image enhancement algorithm has been designed to improve LSM vascular image quality. The main contributions of this thesis are listed as follows: 1. We have conducted a research on mouse brain vessel network imaging method and application with LSM. Combining BABB clearing and immunohistochemical staining we develop a sample process method called iBABB for LSM imaging. Using this method mouse brain vessels endothelium and pericyte is respectively marked by targeting to CD31 and PDGFRβ. Thus we can acquire the mouse brain vessel network with LSM and explore the relationship between endothelium and pericyte of brain vessels. 2. We have designed a stripe artifact elimination method based on non-subsampled contourlet transform (NSCT) for LSM. As regards the problem that stripes drastically deteriorates LSM image quality, an effective stripe artifact elimination method based on non-subsampled contourlet transform (NSCT) and FFT filtering is proposed for both uni-directional and multi-directional LSM. The method is validated by comparative experiment and the result illuminates that our method can well remove stripe noises. 3. We have proposed a vascular image enhancement method. On the unclear imaging of brain micro-vessels with LSM, a L1 norm regularized vessel enhance optimizing model based on gradient modification has been formulated, which is then solved by iterated split bregman algorithm. The method is used to mouse brain vessel network LSM images and dramatically improve the micro-vessel contrast.
关键词	光片显微成像条纹伪影去除血管增强小鼠脑血管成像
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/11819
专题	毕业生_硕士学位论文
作者单位	中国科学院自动化研究所
推荐引用方式 GB/T 7714	梁潇. 脑血管光片显微成像方法与图像处理算法研究[D]. 北京. 中国科学院大学,2016.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
梁潇_脑血管光片显微成像方法与图像处理算（3759KB）	学位论文		限制开放	CC BY-NC-SA