Chinese Optics, Volume. 15, Issue 6, 1211(2022)

Recent progress on the reconstruction algorithms of structured illumination microscopy

Bo ZHOU1, Kun-hao WANG2, and Liang-yi CHEN1,3,4,5、*
Author Affiliations
  • 1Insititute of Molecular Medicine, School of Future Technology, Peking University, Center for Life Sciences United by Peking University-TsingHua University, State Key Laboratory of Membrane Biology, Beijing Key Laboratory of Cardiometabolic Molecular Medicine, Beijing 100871, China
  • 2Key Laboratory of Laser Life Science, Ministry of Education, College of Biophotonics, South China Normal University, Guangzhou 510631, China
  • 3PKU-IDG/McGovern Institute for Brain Research, Beijing 100871, China
  • 4Beijing Academy of Artificial Intelligence, Beijing 100871, China
  • 5National Biomedical Imaging Center, Beijing 100871, China
  • show less
    Figures & Tables(6)
    Schematic diagram of structured illumination microscopy. (a) In sinusoidal illumination microscopy, interference between multiple beams (usually generated by a diffraction grating or spatial light modulator) creates a 2D or 3D striped pattern with spatial frequency illuminating on the sample. This pattern shifts the sample's spatial frequency spectrum to and , translating high-frequency SR information into the diffraction-limited detection passband with the spatial cutoff frequency . After computational processing, the sample's highest detectable frequency can be extended to . (b) Spot-scanning illumination microscopy where fluorescence is collected by an array detector, and pixels offset by a distance from the excitation spot detect a shifted but higher-resolution, low-signal confocal image. The reconstruction algorithm corrects the shift and restores the signal by reassigning the detected fluorescence toward the illumination axis, with the final resolution determined by the product of the excitation PSF () and the emission PSF (). After deconvolution, this process improves resolution similar to that obtained with sinusoidal illumination microscopy
    The schematic diagram of TIRF-SIM (a) and instant SIM (b). Adapted from Kner et al.[35] and York et al.[40]
    Schematic diagram of early implemented 2P SIM (a), 2P-ISIM (b), and 2P SIM with the resonant scanner (c). Adapted from Ingaramo et al.[42], Peter et al.[43] and Gregor et al.[44]
    (a) Schematic diagram of 3D STED-SIM. (b) The cross-section comparison of lateral PSF (top, left), axial PSF (bottom, left), lateral OTF (top, right), and axial OTF (bottom, right) of the widefield microscopy (red) and 3D STED-SIM (blue). Adapted from Xue et al.[49]
    • Table 1. Comparison of SIM SR reconstruction algorithm

      View table
      View in Article

      Table 1. Comparison of SIM SR reconstruction algorithm

      PrincipleEffectCodeReference
      TV-SIMAppend TV regularization to reconstructionSuppress reconstruction artifactsNot open-sourceChu et al. 2014[14]
      Hessian-SIMAppend Hessian regularization to reconstructionSuppress reconstruction artifacts, avoid over-sharpening boundariesOpen-sourceHuang et al. 2018[15]
      HiFi-SIMEngineering the effective SIM PSF into an ideal formSuppress reconstruction artifacts, improve axial sectioningOpen-sourceWen et al. 2021[17]
      Sparse-SIMAppend Sparse and Hessian regularization to reconstructionIncreases SIM resolution ~2-fold laterallyOpen-sourceZhao et al. 2021[26]
      sCMOS Noise-corrected SIMIntroduce sCMOS imaging noise model to reconstructionSuppress sCMOS noise-induced reconstruction artifactsNot open sourceZhou et al. 2022[16]
      Two-step RL deconvolution SIMIntroduce two-step RL deconvolution to reconstructioneliminate ad hoc tuneable parametersNot open sourcePerez et al. 2016[18]
      Noise-controlled SIMIntroduce a physically realistic noise model to reconstructionSuppress reconstruction artifacts, eliminate ad hoc tuneable parameters, maintain resolution and contrastOpen-sourceSmith et al. 2021[19]
      GAN TIRF-SIMUse GAN for transforming TIRF images into TIRF SIM imagesReconstruct rapidlyOpen-sourceWang et al. 2019[20]
      U-Net SIMUse U-net for producing SIM imagesTrain efficiently and reconstruct with fewer low-intensity input imagesOpen-sourceJin et al. 2020[21]
      3D RCANUse 3D RCAN for increasing SIM resolutionIncreases SIM resolution ~1.9-fold laterally and ~3.6-fold axiallyOpen-sourceChen et al. 2021[23]
      DFCAN/DFGANUse DFCAN/DFGAN for producing SIM imagesReconstruct with low SNR input imagesOpen-sourceQiao et al. 2021[24]
    • Table 2. Summary of SIM performance evaluation algorithms

      View table
      View in Article

      Table 2. Summary of SIM performance evaluation algorithms

      FunctionCodeReference
      FRC/FSCDetermine SIM resolution by cross-correlationOpen-sourceNieuwenhuizen et al. 2013[28]
      Decorrelation analysisDetermine SIM resolution by partial phase correlationOpen-sourceDescloux et al. 2019[31]
      NanoJ-SQUIRRELEvaluate SIM artifacts with the resolution scaling functionOpen-sourceCulley et al. 2018[32]
      SIMcheckEvaluate SIM stripe modulation contrast by computing the standard deviationOpen-sourceBall et al. 2015[33]
    Tools

    Get Citation

    Copy Citation Text

    Bo ZHOU, Kun-hao WANG, Liang-yi CHEN. Recent progress on the reconstruction algorithms of structured illumination microscopy[J]. Chinese Optics, 2022, 15(6): 1211

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Review

    Received: Jul. 11, 2022

    Accepted: Aug. 24, 2022

    Published Online: Feb. 9, 2023

    The Author Email: Liang-yi CHEN (lychen@pku.edu.cn)

    DOI:10.37188/CO.EN.2022-0011

    Topics