Deep learning enables stochastic optical reconstruction microscopy-like superresolution image reconstruction from...

Area of Science:

• Biophysics
• Computational Biology
• Microscopy

Background:

• Deep learning shows promise for super-resolution microscopy.
• Challenges exist in achieving high-quality super-resolution from conventional wide-field microscopy.

Purpose of the Study:

• To develop a computational tool for super-resolution microscopy image reconstruction from wide-field images.
• To enable STORM-like super-resolution imaging with input-size flexibility.

Main Methods:

• Developed X-Microscopy, a tool with two deep learning subnets: UR-Net-8 and X-Net.
• Trained models on diverse subcellular structures (cytoskeletal filaments, nanoclusters, etc.).

Main Results:

• X-Microscopy enables STORM-like super-resolution image reconstruction from wide-field microscopy.
• Achieved high-quality reconstructions comparable to STORM-like images.
• Facilitated multicolour and multi-system super-resolution imaging.

Conclusions:

• X-Microscopy enhances super-resolution microscopy capabilities.
• The tool democratizes access to super-resolution imaging beyond specialized labs.