An Algorithm For High-density Localization Of Fluorescent Molecules

The operating principle of localization microscopy is based on repeated cycles ofimaging and localization of fluorescent molecules. Its spatial resolution is determined bythe localization accuracy of fluorescent molecules positioning and the number offluorescent molecules, for improving temporal resolution without sacrificing spatialresolution, It is needed to improve the density of fluorescent molecules at the same timeimproving the localization accuracy. Therefore, development of algorithms forhigh-density localization of fluorescent molecules is of great significance for improvinglocalization microscopy.Existing algorithms for high-density localization of fluorescent molecules researchthe strong signal which is embedded into the weak background, and SSM-BIC algorithmwhich dedicated to the weak signal embedded in strong background choose to extractrectangular region containing the Airy disk, it may also contains gerrymandered signals.From the progress of high-density fluorescence molecules imaging, the complete imageanalysis include sub-region extraction, model selection and fluorescence molecularlocalization algorithm.This paper proposes an algorithm method for high-densitylocalization of weak fluorescent molecules and strong background, called Gaussian fittingalgorithm based on image processing. the region growing method based on the corrosionextracts the irregular region only containing Airy disk，which avoids the interference ofgerrymandered signals.. This algorithm not only improves the sub-region extractionmethod, but also gives the Gaussian fitting model under the irregular region. On the modelselection, compared AIC criterion and minimum error to BIC criterion, BIC criterion onmodel selection performs well on localization. This algorithm considers the accuracy ofeach step, it is found through the empirical analysis that the algorithm on the wholeimproves the effect of localization and positioning accuracy of the high density fluorescentmolecules.