Despite tremendous progress in biological imaging that has yielded tomograms with ever higher resolution, interpretation of data through the segmentation of cell tomograms into subcellular structures (organelles and proteins) remains a challenging task. The difficulty is most extreme in the case of cryo-electron tomography (cryo-ET), where the samples exhibit inherently low contrast due to the limited electron dose that can be applied during imaging before radiation damage occurs, as well as missing-wedge artifacts caused by the limited sample tilt range accessible during imaging. Standard segmentation tools often fail on this type of tomograms. As a result, the segmentation of cellular substructures from a cryo-electron tomogram is currently performed manually in most cases, which is an extremely time-consuming and labor intensive

process.

      In this talk, I will discuss the possibility of using machine learning techniques to segment important subcellular structures. I will examine a strategy that combines convolution neural network, reinforcement learning and classification techniques. The strategy also uses prior knowledge of structure biological to address the ill-posed nature of the segmentation problem. Preliminary results will be presented.

 

海报