Research On Large-scale Localization Data Visualization For Super-resolution Localization Microscopy

Super-Resolution Localization Microscopy(SRLM)is an optical microscopy technique that enables observation of biological structures at the nanometer level.Panoramic superresolution localization microscopy can display the biological structure of an entire sample,helping biologists explore biological tissues and functions,detect events occurred with low probability,and conduct unbiased studies,among others.In order to obtain panoramic images,it is generally necessary to stitch dozens or even hundreds of local field of view localization data into large-scale localization data and visualize it.However,the data volume of large-scale localization data can reach tens or even hundreds of gigabytes,which exceeds the memory of ordinary computers.The existing localization data visualization algorithms can either visualize only small-scale localization data or reduce the image resolution in order to process large-scale localization data,which cannot meet the user’s demand for visualizing panoramic super-resolution images on an ordinary computer.To address the above problems,this paper proposes a large-scale localization data visualization framework,which aims to realize fast interactive visualization of large-scale localization data on an common computer.The details are as follows:(1)For the characteristics of super-resolution localization microscopy data,a multi-resolution hierarchy(Sampling-Level of Detail,sLOD)is constructed from the bottom up based on a random sampling strategy,and a large-scale localization data visualization framework(sLODViewer)is proposed.Specifically,the framework adopts a batch processing technique to divide the large-scale localization data into many small batches,and then the localization data of each small batch is divided into multiple levels according to different accuracies,and dynamically loads the data of the corresponding level according to user requirements.In this way,images that match the size of the computer window can be generated and fast real-time interactive visualization can be achieved.In this paper,we conducted comparison experiments with other visualization methods on the experimental dataset,and the results show that the framework has obvious advantages in processing large-scale localization data,viewing high-resolution images,and fast visualization.(2)To solve the problem of loss of image information and biological structure information caused by the random sampling strategy in the sLODViewer visualization framework,a new lossless visualization framework(fLODViewer)is proposed in this paper.The framework constructs a multi-resolution hierarchy(Fusion-Level of Detail,fLOD)based on the fusion strategy,and invokes and renders fusion points on demand during interactive visualization.A fusion point is a point that fuses all histogram rendering features of localization points in the same spatial region at each resolution level.Based on simulation data and experimental data,this paper demonstrates that the framework is able to generate super-resolution images of the same quality as Ground Truth and retains rich biological structure information while significantly reducing the number of localization points.(3)Design and development of a localization data visualization system.The system uses C++ to implement a large-scale localization data visualization framework for super-resolution localization microscopy,including interactive visualization navigation,biological structure image interception and image multicolor drawing.The system facilitates users to explore the biological characteristics of cellular tissues.