Structured illumination microscopy enhances super-resolution fluorescence imaging but remains sensitive to noise, illumination mismatch and reconstruction artefacts. The authors introduce RL-SIM, a physics-guided reinforcement-learning framework that adaptively optimises structured-illumination parameters during training, improving reconstruction fidelity and robustness in simulations and transferring to fixed-cell and bead experiments on a digital-micromirror-device-based platform.
- Junli Wu
- Qiurong Yan
- Zhiqiang Wen