Abstract

Cellular Electron Cryo-Tomography (CECT) is a powerful imaging technique forthe 3D visualization of cellular structure and organization at submolecularresolution. It enables analyzing the native structures of macromolecularcomplexes and their spatial organization inside single cells. However, due tothe high degree of structural complexity and practical imaging limitations,systematic macromolecular structural recovery inside CECT images remainschallenging. Particularly, the recovery of a macromolecule is likely to bebiased by its neighbor structures due to the high molecular crowding. To reducethe bias, here we introduce a novel 3D convolutional neural network inspired byFully Convolutional Network and Encoder-Decoder Architecture for the supervisedsegmentation of macromolecules of interest in subtomograms. The tests of ourmodels on realistically simulated CECT data demonstrate that our new approachhas significantly improved segmentation performance compared to our baselineapproach. Also, we demonstrate that the proposed model has generalizationability to segment new structures that do not exist in training data.

 

Introduction (beta)

None

 

Conclusion (beta)

None