A Bayesian Method for Tracing Filamentous Structures in Cryo-Electron Microscopy Images

To reveal the structures and functions of filamentous assemblies using cryo-electron microscopy (Cryo-EM), one of the most crucial yet labor-intensive tasks is filament tracing. Existing tracing algorithms are designed primarily for sparse filament networks, often resulting in missed detection in dense configurations. In this paper, we established a Bayesian framework for filament tracing and proposed an algorithm based on the principle of maximum a posteriori probability (MAP) estimation. Our approach simultaneously considers observed data and the continuous nature of filaments, achieving precise tracing even on challenging datasets. Experimental results on diverse real images demonstrate the robustness of our method. In particular, it achieves superior performance over existing techniques in handling dense filament networks.