Fluorescent microscopy has long been an indispensable tool for biological study. However, its advances suffer from limited spatial resolution known as diffraction limit (~200nm), which imposes challenges in investigating tiny biological components such as organelles, viruses, genes, and protein complexes. Single-molecule localization microscopy (SMLM) can be used to resolve subcellular structures and achieve a tenfold improvement in spatial resolution compared to that obtained by conventional fluorescent microscopy. However, the separation of single-molecule fluorescence events that requires thousands of frames dramatically increases the image acquisition time and phototoxicity. The ineluctable tradeoffs among imaging speed, spatial resolution, and cytotoxicity hinder its application in...[ Read more ]

Fluorescent microscopy has long been an indispensable tool for biological study. However, its advances suffer from limited spatial resolution known as diffraction limit (~200nm), which imposes challenges in investigating tiny biological components such as organelles, viruses, genes, and protein complexes. Single-molecule localization microscopy (SMLM) can be used to resolve subcellular structures and achieve a tenfold improvement in spatial resolution compared to that obtained by conventional fluorescent microscopy. However, the separation of single-molecule fluorescence events that requires thousands of frames dramatically increases the image acquisition time and phototoxicity. The ineluctable tradeoffs among imaging speed, spatial resolution, and cytotoxicity hinder its application in live-cell imaging, impeding the observation of instantaneous intracellular dynamics. Here we combine a deep learning network with stochastic optical reconstruction microscopy (STORM) microscope and develop a single-frame super-resolution microscopy (SFSRM) method which takes advantage of the subpixel edge map as well as the multicomponent optimization strategy to guide the neural network to reconstruct a super-resolution image from a single frame of a diffraction-limited image. SFSRM achieves a comparable reconstruction accuracy and resolution to previous methods relying on multi-frames of single-molecule images to reconstruct a super-resolution image. Besides, by the progressive restoration the signal-to-noise ratio and image resolution, SFSRM can handle widefield images with low signal-to-noise ratio, thereby allowing live-cell super-resolution imaging at a superior spatiotemporal resolution and low invasiveness, enabling the investigation of subtle yet fast subcellular dynamics such as diverse interplays between mitochondria with the endoplasmic reticulum, the vesicle transport along microtubules, and the endosome fusion and fission in live cells for over thousands of time points. Moreover, the high robustness of SFSRM to different microscopes and spectra makes it a useful tool for various imaging systems and applications.