Computational Image Analysis Techniques for Super-resolution Localization Microscopy Computational Image Analysis Techniques for Super-resolution Localization Microscopy

Super-resolution microscopy techniques, which overcome the diffraction limit of optical microscopy, are revolutionizing biological research by providing the ability to localize proteins within cells at nanometer resolution. Super-resolution localization microscopy (SRLM) is one of such techniques. It achieves nanometer resolution by randomly activating separate fluorophores and computationally determining their locations. While discoveries of cellular structures at nanometer resolution made recently demonstrated its potential, SRLM still faces several key technological challenges in image analysis. To overcome these challenges, we developed computational tools for analyzing and enhancing SRLM images to enable quantitative characterization and dynamic visualization of cellular processes. One of the computational tools we developed is an image quality assessment method, which directly quantifies the accuracy or precision of the geometry or structure of the SRLM image objects. Another computational tool we developed is an SRLM image segmentation method, which identifies the boundaries of SRLM image objects by kernel density-based method. We utilize correlative fluorescence and atomic force microscopy images to characterize and validate the accuracy of the segmentation method. We have also developed a new data aggregation method under a sliding-window scheme, which improves the temporal resolution to enhance the capability of SRLM imaging of live cells. Together, these computational image analysis techniques enable fundamental extensions of SRLM, which will be essential for quantitative analysis and characterization of dynamic cellular processes.