Physically grounded deep learning-enabled gold nanoparticle localization and quantification in photonic resonator...

Area of Science:

• Biomedical Engineering
• Molecular Diagnostics
• Optical Microscopy

Background:

• Accurate molecular biomarker detection is crucial for disease diagnostics and research.
• Existing methods often lack digital-resolution sensitivity or require complex sample preparation.
• Gold nanoparticles (AuNPs) are promising molecular tags but require precise detection.

Purpose of the Study:

• To develop a deep learning-based method for digital-resolution detection of biomolecules using AuNPs.
• To integrate this method with Photonic Resonator Absorption Microscopy (PRAM) for enhanced sensitivity and accuracy.
• To overcome limitations of conventional methods in quantifying target molecules.

Main Methods:

• Developed LOCA-PRAM (LOcalization with Context Awareness), a deep learning framework.
• Utilized Photonic Crystal (PC)-AuNP resonant-coupling to boost signal contrast.
• Employed Scanning Electron Microscopy (SEM) to determine the point spread function (PSF) for realistic deep learning training data.
• Benchmarked LOCA-PRAM against SEM-derived ground truth.

Main Results:

• LOCA-PRAM achieves digital-resolution detection of biomolecules using AuNPs.
• The method demonstrates high accuracy and sensitivity, surpassing conventional image processing.
• Reliable AuNP detection and localization were achieved even in high-density samples with overlapping PSFs.
• Sub-pixel resolution and accurate quantification of AuNPs were confirmed.

Conclusions:

• PRAM combined with LOCA-based AuNP digital counting offers real-time, high-precision molecular biomarker detection.
• This advancement significantly enhances digital-resolution biosensing capabilities.
• The technology holds promise for improving biomedical research and diagnostic applications.