Abstract:Significance Owing to its non-invasiveness and high specificity, fluorescence microscopy is widely utilized in biomedical research to investigate the structures and functions of biological systems. Limited by the diffraction of light, the resolution of conventional fluorescence microscopy is similar to 250 nanometer (nm) and similar to 800 nm on the lateral and axial axes, respectively, and it cannot resolve nanostructures beyond this limit. To overcome the resolution limit, many super-resolution fluorescence microscopy techniques have been developed, enabling biologists to record the dynamics of the fine structures of organisms and cells in their active states. This offers the potential to elucidate the crucial details of biological phenomena. Nevertheless, in super-resolution fluorescence microscopy, trade-offs exist between resolution, speed, and imaging depth.Although these trade-offs can be moderated by optimizing the microscopy hardware, certain strict physical limitations cannot be easily overcome. Therefore, enhancing microscopy performance via computational imaging methods is particularly important. For instance,the application of deconvolution algorithms can transcend physical limits without changing the optical hardware, thereby improving the dissection of biological information. Progress This review introduces the technical principles of various deconvolution methods. Deconvolution techniques are applied to four modes of super-resolution fluorescence microscopy: structured illumination microscopy (SIM), image scanning microscopy (ISM), stimulated emission depletion (STED) microscopy, and super-resolution optical fluctuation imaging (SOFI). Various modalities have been used for live cell imaging applications. For example, researchers have designed deconvolution algorithms to eliminate the reconstruction artifacts produced during the reconstruction of SIM and to improve its resolution. Additionally, for SOFI,deconvolution techniques can be applied as pre- or post-processing steps to further enhance the efficiency of utilizing statistical information and to improve resolution. The recently developed advanced deconvolution algorithm, sparse deconvolution, is stable and robust to various noise conditions and can effectively improve the three-dimensional resolution two-fold. Furthermore, it can be combined with different variants of fluorescence microscopy to enhance their contrast and resolution in situ without any changes. Owing to significant advances in the corresponding super-resolution reconstruction techniques, live-cell super-resolution microscopy has been effectively enhanced. In the outlook section, considering the unrolling algorithm as an example, this review discusses the prospects of deconvolution methods based on deep learning. The combination of deep learning algorithms and microscopy imaging techniques may become a future development trend in the field of live-cell super-resolution microscopy. This review briefly describes the Fourier ring correlation (FRC) image resolution measurement method and its application in image reconstruction. Finally, a rolling FRC (rFRC) method is introduced to quantitatively detect the reconstruction uncertainties of super-resolution techniques at the corresponding super-resolution scale. Conclusions and Prospects Owing to hardware limitations, extensive super-resolution microscopy methods have introduced computational steps to achieve the optimal quality of super-resolution imaging. This review can serve as a bridge between the super-resolution microscopy and computation communities to facilitate the application of novel computational techniques toward improved resolution, accuracy, and image processing.

Super Resolution Reconstruction of Fluorescence Microscopy Images by a Convolutional Network with Physical Priors.

Universal and High-Fidelity Resolution Extending for Fluorescence Microscopy Using a Single-Training Physics-Informed Sparse Neural Network

Untrained Neural Network Enabling Fast and Universal Structured-Illumination Microscopy

3D resoution enhancement in saturated competition microscopy

Deep Learning Enhanced Fluorescence Emission Difference Microscopy

Ultra-fast, Universal Super-Resolution Radial Fluctuations (SRRF) Algorithm for Live-Cell Super-Resolution Microscopy.

Rapid and Extreme Low-light Superresolution Imaging Via Artificial Intelligence

X-Microscopy: Multicolor Super Resolution Image Reconstruction from Conventional Microscopy with Deep Learning

Zero-shot learning enables instant denoising and super-resolution in optical fluorescence microscopy

Super-Resolution Structured Illumination Microscopy Reconstruction Using a Least-Squares Solver

Deep learning-based super-resolution fluorescence microscopy on small datasets

Time-lapse Image Super-resolution Neural Network with Reliable Confidence Quantification for Optical Microscopy

Single-frame deep-learning super-resolution microscopy for intracellular dynamics imaging

Super-resolution reconstruction of structured illumination microscopy using deep-learning and sparse deconvolution

Deconvolution in Super-Resolution Fluorescence Microscopy(Invited)

High-axial-resolution Single-Molecule Localization under Dense Excitation with a Multi-Channel Deep U-Net.

Deep Learning for Super-Resolution Localization Microscopy

Wide-Field, High-Resolution Reconstruction in Computational Multi-Aperture Miniscope Using a Fourier Neural Network

Faster super-resolution imaging with auto-correlation two-step deconvolution

Evaluation and development of deep neural networks for image super-resolution in optical microscopy

Progress on Applications of Deep Learning in Super-Resolution Microscopy Imaging