近膜区域荧光显微图像中亚细胞目标的斑点检测、融合事件识别和三维形态重建

发布时间：2018-04-27 04:30

本文选题：多角度TIRF成像 + Glut4囊泡　；参考：《浙江大学》2016年博士论文

【摘要】：随着光学显微技术的进步,全内反射荧光显微镜(total internal reflection fluorescence microscopy, TIRFM)相比于传统的生物医学研究方法(如免疫细胞化学、基因技术、亚细胞分层等)能够实时观察细胞膜附近区域内的亚细胞目标(如转运囊泡、自噬溶酶体、细胞骨架等),并定量分析其结构形态和动态运动特性(如微管结构,胞吞胞吐过程中的囊泡运动、细胞迁移等)。但由于所观察的对象特征复杂、运动模式多变,人工分析海量的TIRFM图像数据不仅十分繁琐,分析结果错误率往往较高。因此,利用计算机图像处理技术对显微图像进行自动分析,并提供客观的定量数据帮助量化并验证所观察到的生命活动就显得尤为重要。细胞膜附近区域葡萄糖转运蛋白(glucose transporter, Glut)囊泡的运动情况和微管细胞骨架的结构形态与葡萄糖转运机制密切相关。针对近膜区域内常见的Glut4囊泡、自噬溶酶体以及微管细胞骨架等亚细胞目标的TIRFM图像,本论文开展如下工作：荧光显微图像降噪,显微图像中特征斑点的检测,Glut4囊泡与细胞质膜融合过程的自动识别,以及利用多角度TIRF成像技术重建微管在近膜区域的三维形态。论文主要成果及创新点包括：(1)在分析荧光显微图像形成的基础上,提出基于小波多尺度求和(wavelet multiscale addition, WMA)的图像降噪方法。不同信噪比的模拟图像实验结果表明相对于线性滤波,高斯平滑滤波以及小波多尺度方差稳定变换算法,WMA算法拥有更优的降噪效果。WMA算法在C2C12骨骼肌细胞Glut4囊泡和微管图像中的降噪效果也得到了验证。(2)建立了根据荧光斑点的亮度均值和方差构建特征空间实现显微图像中的特征斑点检测(feature particle detection, FPD)算法。模拟图像实验结果表明在信噪比很低(SNR=I)时,FPD算法的准确检测率仍能达到87%。在胸主动脉平滑肌细胞自噬溶酶体的检测中,FPD算法优于ImageJ软件自带的Analyze articles (AP)算法,与人工识别效果的拟合度达到93%。(3)基于pH值敏感性荧光蛋白VAMP2-pHluorin标记Glut4囊泡,应用移动平均差分算法结合自适应阈值(中值绝对偏差)能从TIRF图像序列中检测出备选Glut4融合囊泡并确定融合起始帧数。通过对备选Glut4融合囊泡进行逐帧二维Gaussian拟合,得到融合过程中囊泡的参数变化情况,以此判断备选Glut4囊泡融合过程属于完全融合(fullfusion)还是部分融合(partial fusion)。三组已知真实融合信息的3T3-L1脂肪细胞TIRF图像序列用于验证识别算法的有效性。实验结果表明备选Glut4融合囊泡准确检测率达到96.5%,融合过程准确识别率达到84.3%。(4)结合多角度TIRF显微技术得到图像的更多空间深度信息,拟合不同入射角下激光的透射强度,实现了观察目标的三维形态重建。实验结果表明在信噪比为2时,探测深度在300nnm范围内的定位精度能达到40nm。U373细胞中微管的三维形态重建结果与细胞迁移过程中细胞骨架的三维空间结构类似,符合生物学意义。本论文提出的TIRF图像处理技术为定量研究细胞膜附近亚细胞目标的结构形态和动态生理活动提供了客观的数据基础,对细胞近膜区域斑点检测、融合的动态过程识别和三维形态重建具有重要的科学意义。
[Abstract]:With the progress of optical microscopy, total internal reflection fluorescence microscopy (TIRFM) is able to observe subcellular targets (such as transport vesicles, autophagy) in the area near the thin cell membrane in real time compared with traditional biomedical research methods (such as immunocytochemistry, gene technology, subcellular stratification, etc.) Lysosome, cytoskeleton and so on), and quantitative analysis of its structure and dynamic motion characteristics (such as microtubule structure, vesicle movement in the process of endocytosis, cell migration, etc.). However, because of the complex characteristics of the endocytosis, the movement pattern is changeable, it is not only very tedious to analyze the massive TIRFM image data artificially, but the error rate of the analysis results is often high. Therefore, using computer image processing technology to automatically analyze the microscopic images and provide objective quantitative data to help quantify and verify the observed life activities is particularly important. The movement of glucose transporter (Glut) vesicles in the vicinity of the cell membrane and the structure of the microtubule cytoskeleton It is closely related to the glucose transport mechanism. In view of the TIRFM images of the common Glut4 vesicles, autophagy lysosomes and microtubule cytoskeleton in the near membrane region, the following work is carried out in this paper: fluorescence microscopic image denoising, detection of feature spots in microscopic images, and automatic identification of Glut4 vesicles and cell membrane fusion process The main achievements and innovation points of this paper are as follows: (1) on the basis of the analysis of the formation of fluorescence microscopic images, the image denoising method based on the wavelet multiscale addition (WMA) is proposed. The simulation image experimental junction with different signal to noise ratio is presented. The results show that compared with linear filtering, Gauss smoothing filtering and wavelet multiscale variance stability transform algorithm, the WMA algorithm has better noise reduction effect.WMA algorithm in C2C12 skeletal muscle cells Glut4 vesicles and microtubule image denoising effect is also verified. (2) building characteristic space based on the mean and variance of luminance spots. The feature particle detection (FPD) algorithm is implemented in the microscopic image. The simulation image experimental results show that the accurate detection rate of FPD algorithm can still reach 87%. in the detection of autophagic lysosomes in the thoracic aorta smooth muscle cells when the signal to noise ratio is very low (SNR=I), and FPD algorithm is superior to Analyze articles (AP) with the ImageJ software. The fitting degree of the algorithm and the artificial recognition effect is 93%. (3) based on the pH value sensitive fluorescent protein VAMP2-pHluorin to mark the Glut4 vesicles. The mobile average difference algorithm combined with the adaptive threshold (absolute deviation of the median) can detect the alternative Glut4 fusion vesicles from the TIRF image sequence and determine the number of the initial frames of the fusion. Through the fusion of the alternative Glut4 The vesicles are fitted by two dimensional Gaussian frame by frame, and the parameters of the vesicles in the fusion process are obtained to determine whether the alternative Glut4 vesicle fusion process belongs to the complete fusion (fullfusion) or partial fusion (partial fusion). The three group of 3T3-L1 fat fine cell TIRF images, known as the true fusion information, are used to verify the effectiveness of the recognition algorithm. The experimental results show that the accurate detection rate of the selected Glut4 fusion vesicles reaches 96.5%, the accurate recognition rate of the fusion process reaches 84.3%. (4) and the multi angle TIRF microscopy is used to obtain more spatial depth information of the image, fitting the transmission intensity of the laser under different incident angles and realizing the three-dimensional morphological reconstruction of the observation target. The experimental results show that the signal noise is in the signal to noise. At 2, the location precision of the detection depth in the range of 300nnm can reach the three-dimensional structure of the microtubule in 40nm.U373 cells, which is similar to the three-dimensional spatial structure of the cytoskeleton during the cell migration. The TIRF image processing technique proposed in this paper is a quantitative study of the structure of the subcellular target near the cell membrane. The state and dynamic physiological activities provide an objective data basis. It is of great scientific significance for the detection of the spot in the cell near the membrane, the recognition of the dynamic process of fusion and the reconstruction of the three-dimensional shape.

【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】