Design optimization of high-sensitivity PCF-SPR biosensor using machine learning and explainable AI.

Area of Science:

• Photonics and Optics
• Biomedical Engineering
• Artificial Intelligence

Background:

• Photonic crystal fiber based surface plasmon resonance (PCF-SPR) biosensors offer precise detection of refractive index variations.
• Existing methods for sensor optimization can be computationally intensive and time-consuming.

Purpose of the Study:

• To introduce a highly sensitive, low-loss PCF-SPR biosensor for label-free analyte detection.
• To integrate machine learning (ML) and explainable AI (XAI) for accelerated sensor design and optimization.
• To evaluate the biosensor's performance for medical diagnostics and chemical sensing.

Main Methods:

• Design and simulation of a PCF-SPR biosensor.
• Application of ML regression techniques to predict optical properties (effective index, confinement loss, amplitude sensitivity).
• Utilization of Shapley Additive exPlanations (SHAP) for identifying critical design parameters.

Main Results:

• Achieved high performance metrics: wavelength sensitivity (125,000 nm/RIU), amplitude sensitivity (-1422.34 RIU⁻¹), resolution (8×10⁻⁷ RIU), and FOM (2112.15).
• ML models showed high predictive accuracy for key optical properties.
• SHAP analysis identified wavelength, analyte RI, gold thickness, and pitch as crucial design factors.

Conclusions:

• The hybrid ML/XAI approach significantly improves sensor design efficiency and reduces computational cost.
• The proposed PCF-SPR biosensor demonstrates potential for high-precision applications like cancer cell detection and chemical sensing.
• A simple yet effective design combined with AI-driven optimization offers a promising path for advanced biosensing.