Single-image-super-resolution Light-sheet Microscopy For Live-cell Imaging

Fluorescence microscopy,which utilizes fluorescent molecules to visualize specific components of biological specimens,is widely used to address fundamental and applied questions in the subcellar study.Light-sheet fluorescence microscopy provides high optical section capability with high temporal resolution,and low phototoxicity is well suited for live-cell imaging.However,the state-of-art light-sheet still have some drawbacks such as low space duty ratio,high side lobes,and chromatic aberration,which prevent the long-term observation of the subcellular interactions with high axial resolution.In addition,the spatial resolution of fluorescence microscopy is limited by the diffraction of light,restricting the observation of the finer interaction of the subcellar interaction.Therefore,using computational imaging methods to further improve the spatial resolution without sacrificing temporal resolution is also of great research significance.This dissertation thoroughly discussed the limitation of spatiotemporal resolution and phototoxicity in live-cell imaging and provided a solution to break these limitations.The following researches are carried out:Firstly,a scanning Bessel light-sheet fluorescence microscopy with the submicron axial resolution is proposed.Through the accurate synchronization of the Bessel beam and the rolling shutter,the out-of-focus fluorescence excited by the side lobe of the Bessel beam is blocked by the camera,which significantly improves the optical section capability.In the software,a dual-stage processing neuron network is processed for single-image-super-resolution.The combination of scanning Bessel-light sheet and dual-stage processing neuron network has achieved 100×100×300 nm~3 spatial resolution in live cells.Secondly,to further improve the space duty cycle and suppress the sidelobes of the Bessel light-sheet,a scan-free uniform Bessel-type light sheet is generated by projecting a line-shaped excitation light double ring modulation plate at the back focal plane of the illumination objective(DR-SPIM).By delicately designing the dual ring modulation plate,an ultra-thin(450nm)and low sidelobe(33%)Bessel light sheet is produced.Based on the double ring light-sheet microscopy,the ultra-high-speed(up to 10 volume/s)and long-term(up to 265 200frames)three-dimensional imaging of subcellular structure is achieved in living cells.In addition,an isotropic divide stage-to-process neural network(ID-Net)is proposed to further improve the fidelity and resolution of the algorithm.This ID-Net separates a live-cell SR restoration into four sequential tasks:denoising,deblurring,depixelization,and isotropic reconstruction,jointly solving the noise,optical blurring,and sensor pixelization anisotropic problems.The ID-Net has achieved isotropic 100 nm resolution with high fidelity at eight kinds of subcellular structures.Finally,by combining the double ring Bessel light-sheet microscopy with an isotropic divide stage to process convolution neural network(IDDR-SPIM),three-dimensional long-term observations of various interactions among subcellular structures with high spatiotemporal resolution are realized in living cells,and the observed three-dimensional interaction of subcellular organelles is quantitatively analyzed.In summary,this dissertation has developed a three-dimensional super-resolution microscopy for living cell imaging with high spatiotemporal resolution and low phototoxicity,which breaks the trade-off between spatial resolution and temporal resolution inherent in current super-resolution microscopy and provides a powerful observation tool for life science research,especially cell biology.