Machine learning-enhanced Kretschmann configuration plasmonic biosensor for real-time brain tumor biomarker detection

Abstract

Brain tumors present significant diagnostic challenges due to delayed symptom manifestation and inadequate early detection methods. Conventional imaging techniques (MRI, CT) lack molecular specificity, while spectroscopic approaches typically exhibit limited sensitivity. In this study, we developed a plasmonic biosensor utilizing Kretschmann configuration with Au/Ag/WS₂ multilayers and graphene for real-time, label-free brain tumor biomarker detection. The sensor exploits surface plasmon resonance amplification in the infrared spectrum to identify distinct molecular vibrational signatures of tumor-associated biomolecules, including altered proteins and lipids.Numerical simulations demonstrate exceptional performance metrics: sensitivity of 804.020°/RIU, detection limit of 0.003 RIU, and figure of merit of 164.086 RIU⁻¹ across refractive indices spanning 1.3333–1.4833, encompassing varied biomarker concentrations. XGBoost machine learning optimization enhances detection accuracy and reliability. Correlation analyses between predicted and Simulation absorption values yield R² coefficients of 88–94 % across WS₂ thickness variations and Ag/Au layer parameters. Under optimized conditions, maximum R² values achieve 94–100 % for Ag layers and 96–100 % for Au layers, confirming robust predictive capability and superior performance under optimal operating parameters. This integrated plasmonic-ML platform represents a promising advancement toward sensitive, early-stage brain tumor diagnosis.