Robust Fourier ptychographic structured illumination microscopy (FP-SIM) using only 3 patterns

Structured illumination microscopy (SIM) holds promising prospects for live-cell super-resolution imaging thanks to its excellent photon efficiency. However, conventional SIM reconstructions require at least 9 exposures for isotropic lateral super-resolution, the compromised imaging speed caused by which still poses significant challenges for live-cell imaging. Moreover, the reliance on precise illumination parameters that are computationally complex to obtain and the vulnerability to low signal-to-noise ratios (SNRs) render conventional SIM reconstructions susceptible to artifacts that degrade fidelity. In this work, we propose a robust Fourier ptychographic SIM reconstruction framework (FP-SIM), which iteratively recovers high-quality, artifact-free super-resolution target with easy access to reliable illumination parameters using only 3 patterns. We demonstrate that FP-SIM can perform robust super-resolution reconstruction with superior overall performance in terms of imaging speed, reconstruction quality, and noise immunity over conventional 9-frame reconstructions at low SNRs. In particular, the open-source code of FP-SIM and associated datasets are provided. Further combined with algorithm acceleration, we achieve real-time super-resolution imaging with a reconstruction frame rate of about 3.75 Hz for a 1024×1024 region of interest. Faster imaging speed, less photodamage, higher reconstruction quality and better noise immunity make FP-SIM a promising tool for studying protein dynamics in live cells.