Exceeding the limits of 3D fluorescence microscopy using a dual-stage-processing network

• DOI: 10.1364/optica.402046
• Publication Date: 2020-10-19T23:00:05Z
• Authors: Hao Zhang; Yuxuan Zhao; Chunyu Fang; Guo Li; Meng Zhang; Yu-Hui Zhang; Peng Fei
• Keywords: Fluorescence-lifetime imaging microscopy

Although three-dimensional (3D) fluorescence microscopy is an essential tool for life science research, the fundamentally limited optical throughput, as reflected in compromise between speed and resolution, so far prevents further movement towards faster, clearer, higher-throughput applications. We herein report a dual-stage mutual-feedback deep-learning approach that allows gradual reversion of degradation from high-resolution targets to low-resolution images. Using single blurred-and-pixelated 3D image input, our trained network infers output with notably higher resolution improved contrast. The performance better than conventional one-stage approaches. It pushes throughput limit current three ways: reducing acquisition time accurate mapping large organs, breaking diffraction imaging subcellular events faster lower-toxicity measurement, improving temporal capturing instantaneous biological processes. Combining light-sheet microscopy, we demonstrate vessels neurons mouse brain at single-cell 6 min whole brain. also cell organelles beyond 2 Hz volume rate map neuronal activities freely moving C. elegans 30 rate.