Enhanced Nanoparticle Recognition via Deep Learning-Accelerated Plasmonic Sensing.

Area of Science:

• Biophysics
• Microscopy
• Machine Learning

Background:

• Surface plasmon microscopy enables label-free imaging of micro/nanoparticles like viruses and bacteria.
• Manual analysis of microscopy images is time-consuming and error-prone, especially with dense samples or noise.
• Automated analysis methods for surface plasmon microscopy data are currently lacking.

Purpose of the Study:

• To develop an accelerated, automated method for analyzing surface plasmon microscopy images of individual particles.
• To address the challenges of manual analysis in dense or noisy biological samples.
• To enable efficient and accurate particle identification for bio-applications.

Main Methods:

• Developed an accelerated approach combining single-particle interference scattering models with deep learning.
• Created hybrid datasets by merging simulated and experimental particle images.
• Constructed a neural network using the EfficientNet architecture for image classification and particle identification.

Main Results:

• Demonstrated the effectiveness of the deep learning technique in classifying interferometric scattering images.
• Successfully identified multiple particles under noisy experimental conditions.
• Validated the approach's capability for automated particle analysis in complex samples.

Conclusions:

• The novel deep learning method significantly improves the analysis of surface plasmon microscopy data.
• This advancement facilitates practical bio-applications by enabling efficient automated particle analysis.
• The technique offers a robust solution for label-free analysis of bio-particles and bio-molecules.