Simultaneous dual-analyte detection biosensor through a stacking-based ensemble machine learning approach: design and...

Area of Science:

• Photonics and Optics
• Biomedical Engineering
• Machine Learning Applications

Background:

• Surface Plasmon Resonance (SPR) sensors are crucial for detecting changes in refractive index.
• Photonic Crystal Fibers (PCFs) offer unique light-confining properties for enhanced sensing.
• Machine learning (ML) models can optimize complex sensor designs and predict performance.

Purpose of the Study:

• To design and optimize a novel PCF-SPR sensor for a broad refractive index (RI) range (1.15-1.41).
• To leverage a stacking ensemble ML model for accurate sensor performance prediction.
• To evaluate the sensor's sensitivity and potential for multi-analyte biosensing.

Main Methods:

• A stacking ensemble ML model combining Random Forest Regressor, Gradient Boosting Regressor, and Decision Tree Regressor was developed.
• Finite Element Method (FEM) with a Perfectly Matched Layer (PML) boundary condition was used for simulation.
• A 40 nm gold layer was applied to the PCF outer surface for SPR excitation.

Main Results:

• The ML model achieved high accuracy with MSE of 0.3584, MAE of 0.2151, and R² of 0.9769.
• The sensor demonstrated exceptional sensitivity: Amplitude Sensitivity (AS) of -1204 RIU⁻¹ and Wavelength Sensitivity (WS) of 10,000 nm/RIU.
• The sensor design is practical and validated for multi-analyte biosensing applications.

Conclusions:

• The developed stacking ML model effectively designs and optimizes PCF-SPR sensors.
• The proposed sensor exhibits high sensitivity and a wide RI detection range, suitable for various biochemical and biological analytes.
• This approach offers a flexible and powerful tool for designing advanced multi-analyte biosensors.