Sensitivity-enhanced machine learning–assisted terahertz Plasmonic biosensor using hybrid 2D materials for tuberculosis detection

This work introduces, a machine learning-assisted terahertz metasurface biosensor that integrates graphene, gold, MXene and phosphorene within a single hybrid architecture for tuberculosis detection. Unlike conventional single- or dual-material plasmonic sensors, the proposed design exploits multi-material plasmonic hybridization to simultaneously achieve high sensitivity, angular robustness and tunable electromagnetic response. Finite element method (FEM) simulations in COMSOL Multiphysics are used to systematically optimize the sensor by varying the graphene chemical potential (0.1–0.9 eV), incident angle (0°–80°) and geometric parameters. The optimized design achieves a maximum sensitivity of 1000 GHz/RIU, figure of merit (FOM) of 14.286 RIU −1 , quality factor of 10.014 and a detection limit of 0.022 RIU, indicating superior performance compared with conventional THz biosensors. Electromagnetic field analysis reveals strong field confinement and hybridized plasmonic modes within the 0.4–1.8 THz range, with a peak absorption of 76.935% at 80° incidence. A linear resonance frequency–refractive index relationship (R 2 = 0.98098) confirms reliable quantitative sensing. Furthermore, the incorporation of machine learning–assisted analysis, yielding an R 2 exceeding 90%, enhances predictive accuracy and robustness. The proposed architecture demonstrates high angular stability, compactness and tunability, establishing its novelty and suitability for point-of-care TB diagnostics and real-time biomedical sensing.