NOISE REDUCTION AND IMAGE ENHANCEMENT IN CRYO-ELECTRON MICROSCOPY: A COVARIANCE WIENER FILTERING APPROACH

Abstract

This thesis, "Denoising and covariance estimation of single particle cryo-EM images", delves into the critical need to enhance the processing and denoising of cryo-electron microscopy (cryo-EM) images for structural biology. Cryo-EM, honored as the "Method of the Year 2015" by Nature Methods and awarded the Nobel Prize in Chemistry in 2017, has revolutionized the field of structural biology by providing high-resolution imaging of biomolecules without the need for crystallization. However, the inherently low signal-to-noise ratio (SNR) of images, caused by minimal electron doses used to prevent radiation damage and the absence of heavy-metal stains for contrast enhancement, poses significant challenges. Traditional denoising techniques, such as Fourier-based filtering and phase flipping, have limitations in handling the complex noise characteristics inherent to cryo-EM images. This research focuses on comparing the effectiveness of two denoising methods: the traditional Wiener filter and the Covariance Wiener Filter (CWF). The latter, by integrating the covariance matrix of projection images into the denoising process, addresses both denoising and Contrast Transfer Function (CTF) correction, leading to enhanced image quality and structural detail. The study seeks to develop and optimize a CWF technique specifically for cryo-EM images, aiming to enhance image quality by effectively reducing noise and preserving the structural integrity of the depicted biological structures. The work includes characterizing noise in cryo-EM images, optimizing covariance matrix estimation, integrating and evaluating CWF in cryo-EM processing pipelines, and analyzing the impact of CWF on biological interpretations. The potential impact of a successful implementation of the CWF extends beyond cryo-EM, providing broader implications for imaging techniques across multiple disciplines where noise significantly impacts data quality. By enhancing the accuracy and reliability of cryo-EM reconstructions, this research aims to advance the technology, providing a robust framework for future research in image processing within and beyond cryo-EM, revolutionizing the field of structural biology, and opening new avenues for studying biological processes and disease mechanisms.