Quantitative Characterization And Analysis Of Image Resolution In Super-resolution Localization Microscopy

Super-resolution localization microscopy(SRLM)is a mainstream super-resolution optical microscopy technique,and has become an essential tool for studying subcellular structure and function.Resolution is a crucial parameter for evaluating image quality in optical microscopy.However,the image quality of SRLM is limited not only by the performance of the optical system,but also by various factors in sample properties and image processing methods.Therefore,in SRLM,researchers typically use reconstructed super-resolution images to characterize resolution(thus called image resolution).Among existing resolution measurement methods,the Full Width at Half Maximum(FWHM)method has been widely used.However,in SRLM,FWHM resolution can be characterized with various ways,and researchers don’t know which way is the best,or how experimental conditions affect the accuracy of the resolution measurement.Moreover,manual selection of a position where the structures are sparse and narrow is required.In this way,the resolution is easily over-estimated due to the selection of incomplete structures.More importantly,the image quality of SRLM is affected by non-uniform experimental conditions,thus the resolving power in different positions changes.If the maximum resolving power(or the best resolution)is still used to characterize the resolution of a superresoluiton image,it will lead to incomplete evaluation of the image resolution or even misjudgment.Based on the above situations,this thesis quantitatively characterized and analyzed the image resolution of SRLM.The main contents are as follows:(1)Analyzed the impact of several important factors in image resolution.A quantitative study of various factors that affect the image resolution will help researchers optimize image resolution characterization methods.Based on the technical principle of SRLM,this thesis discussed the impact of several important factors(including sample size,fluorescence intensity ratio,signal-to-noise ratio,camera pixelation,image processing,etc.)on FWHM resolution.The main results are as follows: a)When it is required to measure FWHM resolution,the deconvolution method can reduce the impact of sample size on resolution,and thus improves the accuracy of resolution measurement;b)When it is required to measure two-point resolution,it is necessary to consider the impact of the peak-to-valley contrast(Central dip);c)The non-uniformity of image resolution needs to be considered,so that misjudgment of resolution values due to the difference in location selection can be avoided.(2)Developed a method and tool for automatic measurement of the best image resolution.Current FWHM resolution measurement method is time-consuming and laborintensive,and is prone to over-estimation.This thesis solved the problem of resolution overestimation through a systematic investigation and optimization of projection FWHM resolution.And,by analyzing the structural information in a super-resolution image,this thesis finds multiple candidate positions for calculating the best projection FWHM resolution,from which the best position is selected via sorting,and thus realizes automatic calcuclation of the best FWHM resolution.Based on the above studies,an Image J plugin named Lucky Profiler was developed to help researchers automatically measure the best FWHM resolution.The effectiveness of the plugin was verified by simulation and experimental data.(3)Developed a new method for quantitative characterization of image resolution nonuniformity(IRNU).In SRLM,the best resolution is usually used by current resolution measurement methods to characterize the resolution of an entire super-resolution image,and the existence of IRNU is ignored,which can result in an incomplete resolution evaluation or even misjudgment.This thesis proposed a new concept called statistical resolution,and developed the corresponding Double fitting method(which is based on FWHM resolution)for quantitative characterization and analysis of IRNU.The study found that IRNU obeys a log-normal distribution,which can be characterized by Gaussian model,and can be quantified by three statistical parameters.Using these parameters,this thesis compared the performance of different localization algorithms and different super-resolution optical microscopy.In summary,to address the need for quantitative characterization of image resolution in SRLM,this thesis studied the influencing factors of image resolution,developed an automatic and accurate tool for image resolution measurement,and proposed a new method to quantitatively characterize IRNU.This thesis helps researchers understand and accurately characterize the image resolution in SRLM,and thus facilitate the development and promotion of SRLM.