Modified K-factor image decomposition for three-dimensional super resolution microscopy

The ability to track single fluorescent particles within a three dimensional (3D) cellular environment can provide valuable insights into cellular processes. In this paper, we present a modified nonlinear image decomposition technique called K-factor that reshapes the 3D point spread function (PSF) of an XYZ image stack into a narrow Gaussian profile. The method increases localization accuracy by ∼60% with compare to regular Gaussian fitting, and improves minimal resolvable distance between overlapping PSFs by ∼50%. The algorithm was tested both on simulated data and experimentally. This work presets a novel use of the nonlinear image decomposition technique called K-factor that reshapes the three dimensional (3D) point spread function (PSF) of an XYZ image stack into a narrow Gaussian profile. The experimentally obtained PSF of a Z-stack raw data that is acquired by a widefield microscope has a more elaborate shape that is given by the Gibson and Lanni model. This shape increases the computational complexity associated with the localization routine, when used in localization microscopy techniques. Furthermore, due to its nature, this PSF spreads over a larger volume, making the problem of overlapping emitters detection more pronounced. The ability to use Gaussian fitting with high accuracy on 3D data can facilitate the computational complexity, hence reduce the processing time required for the generation of the 3D superresolved image. In addition it allows the detection of overlapping PSFs and reduces the effects of the penetration of out of focus PSFs into in focused PSFs, therefore enables the increase in the activated fluorophore density by ∼50%.