基于压缩感知的三维单分子定位显微成像方法研究  

Three-dimensional single-molecule localization microscopy imaging based on compressed sensing

作　　者：张赛文 林丹樱[3] 于斌[3] 冷潇泠 张光富 田野[1,2] 谭伟石 ZHANG Sai-wen;LIN Dan-ying;YU Bin;LENG Xiao-ling;ZHANG Guang-fu;TIAN Ye;TAN Wei-shi(School of Information and Electronics Engineering,Hunan City University,Yiyang 413000,China;All-solid-state Energy Storage Materials and Devices Key Laboratory of Hunan Province,Hunan City University,Yiyang 413000,China;Key Laboratory of Optoelectronic Devices and Systems of Ministry and Education and Guangdong Province,College of Physics and Optoelectronics Engineering,Shenzhen University,Shenzhen 518060,China)

机构地区：[1]湖南城市学院信息与电子工程学院,湖南益阳413000 [2]湖南城市学院全固态储能材料与器件湖南省重点实验室,湖南益阳413000 [3]深圳大学物理与光电工程学院,光电子器件与系统(教育部/广东省)重点实验室,深圳518060

出　　处：《中国光学》2020年第5期1065-1074,共10页Chinese Optics

基　　金：国家自然科学基金资助项目(No.11947088,No.11604091,No.11547186,No.61775144,No.61975131);湖南省自然科学基金项目(No.2019jj50025,No.2018JJ2019);湖南省教育厅科学研究项目优秀青年项目(No.19B100,No.19B098)。

摘　　要：本文建立了一种三维压缩感知模型以实现对高密度荧光分子图像的快速三维定位。首先,根据荧光显微的三维点扩展函数成像理论,设计测量矩阵,并建立压缩感知模型。接着,对荧光显微成像过程进行了模拟,并采用凸优化方法(CVX)、正交匹配追踪(Orthogonal Matching Pursuit,OMP)算法和同伦算法对建立的压缩感知模型中模拟生成的图像进行了定位分析,分别从恢复率、定位精度、重构时间几方面进行了对比。最后,采用同伦算法对模拟的生物样品和实验室采集的细胞进行了三维定位,并获得了三维超分辨图像。对比结果表明:在重构密度和定位精度接近的情况下,同伦算法比CVX方法的重构速度快2个数量级。同伦算法较OMP算法的定位精度要高一倍。采用同伦算法来实现三维的超分辨荧光显微成像在节约计算时间、实现实时成像方面具有一定的意义。In order to achieve fast three-dimensional localization of high-density fluorescent molecular images,a three-dimensional compressed sensing model was established and studied using the CVX method,the Orthogonal Matching Pursuit(OMP)algorithm and a homotopy algorithm.The models’measurement matrix was then designed.Firstly,the system’s theory and design were both developed using the three-dimensional point-spread function imaging theory of fluorescence microscopy.Then,the process of fluorescence microscopic imaging was simulated,through which the images generated in the established compressed sensing model were analyzed using the CVX method,OMP algorithm and homotopy algorithm.The recall rate,localization accuracy and reconstruction time were compared.Finally,the simulated biological samples and the collected cells in the laboratory were analyzed using the homotopy algorithm,and thus three-dimensional super-resolution imaging was achieved.It can be seen from the comparative results that the homotopy algorithm is two orders of magnitude faster than the CVX method when the reconstruction density and localization accuracy have little deviation.The localization accuracy of the homotopy algorithm is twice higher than that of the OMP algorithm.The homotopy algorithm is meaningful for 3D super-resolution fluorescence microscopy imaging,which can save computing time and achieve real-time imaging.

关 键 词：压缩感知 单分子定位 图像重构 同伦算法 

分 类 号：O439[机械工程—光学工程]