Conditional diffusion model to enhance optical sectioning microscopy

Structured illumination microscopy (SIM) yields optically-sectioned images by projecting fringe patterns on a sample and laterally shifting the fringes three times. In this paper, we introduce a conditional framework entitled optically-sectioning diffusion (OSdiffuse) model for 3D microscopic imaging. This framework reconstructs a sectioned image using a single wide-field image as input. It effectively suppresses background noise similar to optical sectioning SIM (OS-SIM), but it reduces the data acquisition by threefold. Furthermore, the model enhances the axial sectioning capability twofold compared to the conventional OS-SIM approach. The effectiveness of the proposed mode was demonstrated with both simulations and experiments. We believe that the proposed method will be widely used in biological studies.