A Single-Shot Autofocus Approach for Surface Plasmon Resonance Microscopy.

Area of Science:

• Microscopy
• Nanotechnology
• Artificial Intelligence

Background:

• Surface Plasmon Resonance Microscopy (SPRM) is a sensitive imaging technique for chemical and biological analysis.
• SPRM systems face challenges with focus inhomogeneity and drifts, impacting image quality and quantification, especially during long-term recordings.
• Existing focus correction methods often require complex optical modifications.

Purpose of the Study:

• To develop a deep-learning-based image processing method for autofocusing SPRM images.
• To achieve autofocused SPRM images without altering the optical system's complexity.
• To enhance the accuracy and consistency of SPRM imaging for various applications.

Main Methods:

• A generative adversarial network (GAN) model was trained using a large dataset of nanoparticle SPRM images captured at varying focal distances.
• The trained GAN model was designed to generate focused images directly from single-shot, defocused SPRM images.
• The method operates without requiring prior knowledge of the focal conditions during image acquisition.

Main Results:

• The proposed deep-learning method successfully generated autofocused SPRM images from defocused inputs.
• Experiments with gold nanoparticles demonstrated the effectiveness of the autofocus technique in both static and time-lapse imaging scenarios.
• The method showed significant potential for improving image quality and data reliability in SPRM studies.

Conclusions:

• The developed deep-learning autofocus technique offers a non-invasive solution to SPRM focus issues.
• This approach enhances the consistency and reliability of SPRM for long-term monitoring and quantitative analysis.
• The method paves the way for more robust and accessible SPRM applications in scientific research.