Algorithmic corrections for localization microscopy with sCMOS cameras - characterisation of a computationally efficient localization approach

Abstract

Modern sCMOS cameras are attractive for single molecule localization microscopy (SMLM) due to their high speed but suffer from pixel non-uniformities that can affect localization precision and accuracy. We present a simplified sCMOS non-uniform noise model that incorporates pixel specific read-noise, offset and sensitivity variation. Using this model we develop a new weighted least squared (WLS) fitting method designed to remove the effect of sCMOS pixel non-uniformities. Simulations with the sCMOS noise model, performed to test under which conditions sCMOS specific localization corrections are required, suggested that pixel specific offsets should always be removed. In many applications with thick biological samples photon fluxes are sufficiently high that corrections of read-noise and sensitivity correction may be neglected. When correction is required, e.g. during fast imaging in thin samples, our WLS fit procedure recovered the performance of an equivalent sensor with uniform pixel properties and the fit estimates also attained the Cramer-Rao lower bound. Experiments with sub-resolution beads and a DNA origami test sample confirmed the results of the simulations. The WLS localization procedure is fast to converge, compatible with 2D, 3D and multi-emitter localization and thus provides a computationally efficient sCMOS localization approach compatible with most SMLM modalities.