Solving the edge artifact in the SUPPOSe algorithm for super-resolution of fluorescent images

SANDRA MARTINEZ; TOSCANI, MICAELA; MARTINEZ, OSCAR E.

Congreso; XLVIII Reunión Anual SAB; 2019

Trying to apply SUPPOSe (a super-resolution deconvolution algorithm [1]) to the axonalcytoskeleton (for the actin/spectin) to detect the presence of a periodic structure ofperiodicity ≈ 180 nm − 190 nm denominated membrane-associated periodic skeleton(MPS) [2,3], we found the appearance of an artifact in the reconstruction that hinderedan appropriate recovery of the structure.It was determined that the artifact arises from the presence of fluorescent moleculesoutside the registered frame but close enough to the boundary contribute withsignificant signal inside the frame.The algorithm presented in [1] for wide-field fluorescent microscopy deconvolution fromsingle acquisition consists on assuming that the measure image is a superposition ofpoint sources of equal intensity. The goal of this method was to find the positions of suchsources. This optimization problem was solved using a genetic algorithm with anappropriated fitting function. In this new work a new fitting function is used to removeartifacts arising from the edges.To test the improved algorithm with an image with known ground truth we constructed asynthetic image similar to MPS consisting of k parallel lines with different periods p. Thepoint spread function (PSF) is Gaussian similar to the one obtained with a wide-fieldfluorescent microscopy, for this special set-up the PSFwidth=95nm.References1. S. Martínez, M. Toscani, O. E. Martínez, https://doi.org/10.1111/jmi.12802. 07 May2019.2. K. Xu, G. Zhong, and X. Zhuang, Science, vol. 339,no. 6118, pp. 452?456, 2013.3. N. Unsain, et al., Scientific reports, vol. 8, no. 1, p. 3007, 2018