Diabetes management requires precise and frequent glucose monitoring; however, existing techniques remain invasive, painful and insufficient for continuous long-term use, limiting patient compliance and accessibility. To address these limitations, a surface plasmon resonance (SPR)-based glucose biosensor incorporating machine-learning-optimized black phosphorus (BP) sensing layers is proposed. A detailed analysis of the sensor design has been conducted using Maxwell’s Equations and Transfer Matrix Method (TMM) in order to optimize the thickness of the Au (7-54 nm), BP (0.2-2.2 nm) and SrTiO 3 (0.3-2.3 nm) layers followed by numerical validation using COMSOL Multiphysics. The optimized structure exhibits minimum reflectance values ranging from 0.253 % to 0.955 % for glucose-induced refractive index variations, corresponding to resonance angle shifts from 74° to 76.2°. A maximum local sensitivity of 300°/RIU is achieved over a narrow refractive index interval, while the overall sensitivity across the full sensing range varies between 166°/RIU and 183°/RIU. This performance surpasses many existing SPR sensors while maintaining a figure of merit of 76 and a detection accuracy of 0.152. A strong linear correlation between resonance angle and refractive index (R 2 = 0.99734), expressed as θ(°) = 178.5714RI − 164.3405, confirms excellent sensing precision. Furthermore, machine learning regression models demonstrate robust predictive performance with R 2 values ranging from 0.92 to 1.00, significantly enhancing real-time glucose response estimation. Electric field distribution analysis reveals maximum field confinement at the metal-dielectric interface at a 75° incident angle, ensuring efficient analyte interaction. These results demonstrate that the proposed SPR biosensor is highly sensitive, accurate, and suitable for intelligent wearable sensing applications for next-generation non-invasive glucose monitoring and diagnostics.
Ponlatha et al. (2026) studied this question.