Equivariance-Based Theoretical Analysis and Self-Supervised Learning Framework for Missing Wedge Problem in Cryo-Electron Tomography

Cryo-electron tomography (cryo-ET) is an advanced imaging technique for studying biological macromolecules within their physiological environment, providing valuable insights into macromolecule structures and functions. However, due to limitations in image acquisition, cryo-ET suffers from the missing wedge problem, which leads to image distortions and artifacts that complicate subsequent analysis. In this study, we utilize the prior information inherent in cryo-ET data to undertake a theoretical analysis of the missing wedge problem and establish the conditions for signal recovery. We propose an enhanced self-supervised learning framework with a full angular space rotation scheme to address the missing wedge problem. The effectiveness of the proposed methods is validated on datasets of varying content complexity and compared with the state-of-the-art methods, demonstrating significant improvements in image quality and the preservation of structural integrity.