Fast adaptive super-resolution lattice light-sheet microscopy for rapid, long-term, near-isotropic subcellular imaging

Abstract

Lattice light-sheet microscopy (LLSM) provides a crucial observation window into intra- and inter-cellular physiology of living specimens with high speed and low phototoxicity, however, at the diffraction-limited resolution or anisotropic super-resolution with structured illumination. Here we present the meta-learning-empowered reflective lattice light-sheet virtual structured illumination microscopy (Meta-rLLS-VSIM), which instantly upgrades LLSM to a near-isotropic super resolution of [~]120-nm laterally and [~]160-nm axially, more than twofold improvement in each dimension, without any modification of the optical system or sacrifice of other imaging metrics. Moreover, to alleviate the tremendous demands on training data and time necessitated by existing deep-learning (DL) methods, we devised an adaptive online training approach by synergizing the front-end imaging system and back-end meta-learning framework, which reduced the total time for data acquisition and model training down to tens of seconds. With this method, a new model can be well-trained with tenfold less data and three orders of magnitude less time than current standard supervised learning. We demonstrate the versatile functionalities of Meta-rLLS-VSIM by imaging a variety of bioprocesses with ultrahigh spatiotemporal resolution for long duration of hundreds of multi-color volumes, characterizing the dynamic regulation of contractile ring filaments during mitosis and the growth of pollen tubes, and delineating the nanoscale distributions, dispersion, and interaction pattern of multiple organelles in embryos and eukaryotic cells.