Fast imaging of 15 intracellular compartments and interactions by deep learning segmentation of super-resolution data

Karl Zhanghao,Meiqi Li,Xingye Chen,Wenhui Liu,Tianling Li,Yiming Wang,Fei Su,Zihan Wu,Chunyan Shan,Jiamin Wu,Yan Zhang,Jingyan Fu,Peng Xi,Dayong Jin

DOI: https://doi.org/10.21203/rs.3.rs-1686675/v1

2022-01-01

Abstract:Abstract The number of colors in fluorescence microscopy is far less than the types of intracellular compartments, bringing gaps in studying live-cell anatomy and multiple organelles’ interactions. Here, we report that super-resolution imaging in association with deep convolutional neuronal networks can predict 15 subcellular structures at > 91.7% pixel accuracy using one laser excitation and two detection channels. It not only bypasses the limitations of multi-color imaging with single dye labeling but also accelerates the imaging speed by more than one order of magnitude. We find that the super-resolution ratiometric images well reflect the heterogeneity of organelles as the intrinsic “optical fingerprint” and the neuronal networks can be generalized with transfer learning to predict both 3D and 2D datasets from different microscopes, different cell types, and even complexed system of living tissues. It enables us to resolve the 3D anatomic structure of live cells at different mitotic phases and to track down the fast dynamic interactions among 9 intracellular compartments.