Deep Learning Approach for Enhanced Detection of Surface Plasmon Scattering

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Biomedical Imaging

Background:

• Surface plasmon microscopy (SPM) is crucial for analyzing light scattering.
• Conventional detection methods in SPM face limitations in noisy environments and with complex scattering parameters.
• Improving detection accuracy is vital for advanced microscopy and molecular detection.

Purpose of the Study:

• To develop and assess a deep learning approach for enhancing SPM detection characteristics.
• To quantitatively evaluate the performance improvement of deep learning over conventional SPM detection.
• To explore the applicability of deep learning in label-free molecular detection assays.

Main Methods:

• Utilized a deep learning approach based on convolutional neural network (CNN) algorithms.
• Applied the deep learning model to estimate scattering parameters, focusing on the number of scatterers.
• Assessed the improvement quantitatively using SPM images formed by coherent interference of scatterers.

Main Results:

• Deep learning significantly improves detection accuracy compared to conventional methods.
• Accuracy enhancement was found to be approximately six times higher.
• The approach demonstrated effectiveness in analyzing scattering from polydisperse mixtures.

Conclusions:

• Deep learning offers a powerful tool for effectively detecting scattering objects in noisy environments.
• The developed deep learning method substantially enhances SPM performance.
• This approach holds promise for advancing label-free molecular detection and other imaging techniques.