Related Experiment Video
Updated: May 2, 2026
08:04
Multifunctional Hybrid Fe2O3-Au Nanoparticles for Efficient Plasmonic Heating
Published on: February 20, 2016
14.3K
Machine learning-driven optimization of monolithic gold plasmonic sensors: Achieving ultrahigh sensitivity with
Sonia Akter1, Hasan Abdullah1,2
1Bangladesh Army International University of Science and Technology (BAIUST), Cumilla, Bangladesh.
Plos One
|March 13, 2026
Summary
This study combines machine learning with photonic crystal fiber sensors for advanced biosensing. Optimized gold-coated sensors and ML algorithms achieve high sensitivity and accuracy in detecting biological analytes.
Area of Science:
- Nanophotonics
- Biosensing
- Machine Learning
Background:
- Surface plasmon resonance (SPR) biosensing offers high sensitivity but faces challenges in performance optimization and data interpretation.
- Integrating advanced computational frameworks with novel sensor designs is crucial for enhancing biosensing capabilities.
Purpose of the Study:
- To develop a highly sensitive and accurate SPR biosensor by co-designing a gold-coated photonic crystal fiber (PCF-SPR) sensor with a machine learning (ML) computational framework.
- To evaluate the performance of different ML regression models for predicting the optical responses of the PCF-SPR sensor.
Main Methods:
- An asymmetric circular photonic crystal fiber (PCF) with a 50 nm gold coating was designed to maximize evanescent field-analyte interaction.
- COMSOL Multiphysics simulations generated synthetic data for refractive index variations (1.33-1.38), capturing key optical parameters.
- Multiple Linear Regression (MLR), Support Vector Regression (SVR), and Random Forest Regression (RFR) models were employed and compared for predictive accuracy.
Main Results:
- The PCF-SPR sensor achieved a record wavelength sensitivity of 31,846.46 nm/RIU with minimal variation (0.02%) across the biological refractive index range.
- A resolution of [Formula: see text] RIU was demonstrated, indicating high precision in analyte detection.
- Multiple Linear Regression (MLR) outperformed nonlinear models, showing superior accuracy in predicting confinement loss (MAE = 3.97, RMSE = 5.03) and sensitivity (MAE = 40.18, RMSE = 50.54).
Conclusions:
- The synergistic integration of optimized gold-microstructures and interpretable machine learning provides a robust pipeline for high-sensitivity, noise-resilient biosensing.
- This approach surpasses previous ML-enhanced plasmonic sensors in key performance metrics while simplifying sensor fabrication.
- The developed framework establishes a new benchmark for advanced biosensing applications.
