Enhanced 2D image reconstruction for 3D structured illumination microscopy using the plug-and-play priors framework

Area of Science:

• Microscopy
• Image Reconstruction
• Computational Imaging

Background:

• Structured illumination microscopy (SIM) is crucial for high-resolution imaging.
• Existing SIM reconstruction methods often struggle with out-of-focus light and computational cost.
• 3D SIM requires significant computational resources for reconstruction.

Purpose of the Study:

• To develop an efficient reconstruction method for 3D structured illumination microscopy.
• To suppress out-of-focus light effectively without full 3D computation.
• To improve image quality metrics like PSNR and SSIM in 3D SIM.

Main Methods:

• Integration of a pretrained denoising neural network as a plug-and-play (PnP) prior.
• Application within a physics-based Proximal ADMM framework.
• Combination of thick-slice 2D processing with a DnCNN prior and 3D point spread function incorporation.

Main Results:

• PnP-SIM enables efficient reconstruction of selected focal planes.
• The method effectively suppresses out-of-focus light.
• Demonstrated superior performance over 2D-FairSIM in PSNR and SSIM on simulated and experimental data.
• Achieved performance comparable to computationally intensive 3D reconstruction methods.

Conclusions:

• PnP-SIM offers an efficient and effective solution for 3D SIM reconstruction.
• The method balances computational efficiency with high-quality image reconstruction.
• PnP-SIM represents a significant advancement in 3D structured illumination microscopy.