Cryo-Electron Microscopy Image Registration Using Projection Method for Fourier-Bessel Transform Computation

Abstract

Projection alignment in single particle cryo-electron microscopy is an important step of obtaining a high-resolution three-dimensional model. To solve this problem, one needs a method that can quickly process thousands of projection images. In this paper, we propose a single particle cryo- electron microscopy image registration approach based on the calculation of the Fourier-Bessel transform using the projection method, namely, the expansion of the Hankel transform into a series of its eigen functions - the Laguerre functions. The proposed projection method significantly speeds up the calculation of the registration parameters. The proposed approach was evaluated on the real and synthetic data with different level of noise and compared to the existing approaches. The experiments showed that the proposed approach has a high registration quality especially for noisy data. In addition, the proposed algorithm significantly accelerates the computations compared to the original registration method based on Fourier-Bessel transform.