High Resolution of Plasmonic Resonance Scattering Imaging with Deep Learning

Area of Science:

• Optical microscopy
• Nanotechnology
• Computational imaging

Background:

• Dark-field microscopy (DFM) offers high signal-to-noise for nanoscale imaging.
• The optical diffraction limit restricts resolution, challenging the distinction of closely spaced nanoparticles.
• Accurate imaging of nanoscale structures remains a significant challenge in optical microscopy.

Purpose of the Study:

• To develop a computational strategy for enhancing DFM image resolution.
• To overcome the diffraction limit in DFM for improved nanoscale imaging.
• To enable high-resolution imaging of plasmonic scattering and biological samples.

Main Methods:

• A deep learning framework, NanoNet, was developed for image segmentation.
• NanoNet was trained using paired DFM and scanning electron microscopy (SEM) images.
• The method segments scattered light spots in diffraction-limited DFM images.

Main Results:

• NanoNet successfully generated high-resolution DFM images from low-resolution inputs.
• The achieved resolution matched that of scanning electron microscopy (SEM).
• A high-resolution DFM image of living cells was successfully obtained without complex optical setups.

Conclusions:

• The NanoNet deep learning approach effectively enhances DFM resolution beyond the diffraction limit.
• This computational strategy provides a new pathway for high-resolution optical nanoscopic imaging.
• The technique allows for detailed visualization of nanoscale phenomena, including biological samples.