Font Size: a A A

Spots Detection, Fusion Events Identification And 3D Structures Reconstruction Of Subcellular Objects In Fluorescence Microscopy Images Near The Plasma Membrane

Posted on:2017-04-28Degree:DoctorType:Dissertation
Country:ChinaCandidate:J WuFull Text:PDF
GTID:1108330485957084Subject:Biomedical engineering
Abstract/Summary:PDF Full Text Request
With the development of new optical microscopy, total internal reflection fluorescence microscopy (TIRFM) compared to some traditional biomedical research methods (e.g. immunocytochemistry, gene technology, subcellular isolation) can observe subcellular objects near the cell membrane in real time (e.g. transporter vesicle, autolysosome, cytoskeleton), and quantitatively analyze their dynamic activities (e.g. microtubule structures, vesicle movement in endocytosis or exocytosis, cell migration). Due to the complex features of subcellular objects and subtle movement patterns, manually analyzing massive TIRF images is not only very complicated, always results in error. It is particularly important to use computer image processing for automatic analysis of microscopic image and provide objective quantitative data to help quantify and verify the observed activities of life.The movement of glucose transporter (Glut) vesicles and microtubule structures near the cell membrane are closely related to glucose transportation mechanism. Based on TIRFM images of common subcellular objects, such as Glut4 vesicles, autolysosomes, microtubules, some works carried out in this thesis are as follows:denoising fluorescence microscopy images, detecting feature particles in fluorescence microscopy images, automatically identifying fusion events between Glut4 vesicles and the plasam membrane, and three-dimensional (3-D) reconstruction of microtubules near the cell membrane with multi-angle TIRF images. The main innovations of this thesis include:(1) Through analyzing the process of fluorescence microscopy imaging, a denoising algorithm based on wavelet multiscale addition (WMA) was proposed. The validation results with simulated images at different noise level show that WMA comparied to linear smoothing, Gaussian smoothing and wavelet multiscale variance stabilizing transform (WMVST) can achieve better results. The denoising results of WMA, which applied to Glut4 vesicle and microtubule images in C2C12 skeletal muscle cells, have also been verified.(2) According to the mean and variance of the fluorescent spot intensities, the feature space can be constructed to achieve feature particle detection (FPD). The validation results with simulated images show:The F-score of FPD algorithm can reach 87% at the noise level SNR=1. During the autolysosome detection in thoracic aortic smooth muscle cell, FPD algorithm can surpass Analyze articles (AP) algorithm in ImageJ software and fit the ground truth with R2=93%.(3) Based on Glut4 vesicles labeled the pH-sensitive fluorescence protein VAMP2-pHluorin, the moving average difference algorithm combined with an adaptive threshold (median absolute deviation, MAD) can detect Glut4 fusion candidates and identify the frame of fusion start in TIRF image sequences. Through two-demensional (2-D) Gaussian fitting of each frame in a Glut4 fusion candidate, some parameters can be derived during the fusion process and used to determine the fusion type:full fusion or partial fusion. Three TIRF image sequences of 3T3-L1 adipocytes with known ground truth were used to verify the algorithm. Experimental results show that the detection rate of Glut4 fusion candidate can reach 96.5%, while the the identification rate of fusion event can reach 84.3%.(4) Multi-angle TIRF images contain more spatial depth information and can be used to reconstruct three-dimensional (3-D) structures by fitting the transmission intensity of the laser at different incidence angles. Simulation results show that the location resolution of our method at the low noise level SNR=2 can achieve 40nm within 300nm penetration depth. The 3-D structures of microtubules derived from our method match a similar 3-D structure of the cytoskeleton during the cell migration in U373 cells.The TIRF image processing algorithms proposed in this thesis can lay a solid foundation for the quantitative study of subcellular objects near the cell membrane, and has important scientific significance in spot detection, fusion events identification and three-dimensional reconstruction.
Keywords/Search Tags:multi-angle TIRF imaging, Glut4 storage vesile, autolysosome, microtubule, denoising, feature particle detection, fusion identification, 3-D reconstruction
PDF Full Text Request
Related items