DHPSFU: a Fiji plugin for fast and accurate double helix-PSF 3D single-molecule localisation microscopy

Abstract

Abstract The double-helix point-spread function (DH-PSF) is one of the most used PSFs for large depth-of-field 3D single-molecule localisation microscopy. Due to its popularity, many algorithms have been developed to analyse experimental DH-PSF data, either based on dedicated DH-PSF fitting or on generalised PSF fitting, typically using cubic splines. We show here that the most popular implementations of both these approaches have limitations in terms of localisation performance, processing speed or user-friendliness. To overcome some of these limitations, we have developed a new analytical approach for DH-PSF fitting based on unmixing (DHPSFU) of fitted localisation data using distance pairing. We compare DHPSFU with two popular algorithms, SMAP and EasyDHPSF, using realistic simulated datasets based on experimental data, to show that our algorithm achieves the highest Jaccard index (DHPSFU: 0.98; SMAP: 0.91; EasyDHPSF: 0.85) and fastest CPU-based processing speed (DHPSFU: 6,800 locs/s; SMAP: 2,500 locs/s; EasyDHPSF: 63 locs/s). We also show that our algorithm achieves the best resolution when imaging the cellular plasma membrane of Jurkat T cells (DHPSFU: 140 nm, EasyDHPSF: 162 nm, SMAP: 165 nm). We have incorporated DHPSFU as a Fiji plugin and provide Matlab and Python scripts for user customisation.