Multi-task Learning for Macromolecule Classification, Segmentation and Coarse Structural Recovery in Cryo-Tomography

摘要: Cellular Electron Cryo-Tomography (CECT) is a powerful 3D imaging tool for studying the native structure and organization of macromolecules inside single cells. For systematic recognition recovery macromolecular structures captured by CECT, methods several important tasks such as subtomogram classification semantic segmentation have been developed. However, are still very difficult due to high molecular structural diversity, crowding environment, limitations CECT. In this paper, we propose novel multi-task convolutional neural network model simultaneous classification, segmentation, coarse interest in subtomograms. our model, learned image features one task shared thereby mutually reinforce learning other tasks. Evaluated on realistically simulated experimental CECT data, outperformed all single-task segmentation. addition, demonstrate that can generalize discover, segment recover do not exist training data.