Inverse modelling of point spread function for single-molecule localization microscopy.

Abstract

Single-molecule localization microscopy (SMLM) is one major type of super-resolution microscopy. SMLM is based on localizing single-molecule emission patterns, i.e. the point spread function (PSF) at a precision much smaller than the diffraction limit. PSF is the impulse response of an imaging system, it contains information of the image formation, systematic aberrations and imperfections. PSF modelling is essential in SMLM to achieve both high accuracy and precision. Here I present a universal PSF learning algorithm to reconstruct the PSF models from fluorescence bead data. It can extract the PSF models from single and multi-channel systems, learn the field dependent aberration and model the depth dependent aberrations. The learned PSF model can be in forms of 3D matrices, Zernike coefficients or image-based pupil functions according to user selections. The learning framework can be also extended to in-situ PSF learning, where the PSF models are directly extracted from single-molecule blinking data from SMLM imaging. I demonstrate in-situ PSF learning from single-channel systems with various aberrations.