A High Sensitivity Terahertz Biosensor with Hybrid Metasurfaces for Tuberculosis Detection Leveraging the Integration of Machine Learning and Multi-Material Resonators

This study presents a terahertz-based biosensor for tuberculosis detection, incorporating a unique metasurfaces configuration. The sensor’s architecture features multiple resonating elements: a silver-based circular ring resonator, a black phosphorus square ring structure, a graphene platform, and symmetrically positioned gold rectangular resonators. Finite element method analysis through COMSOL Multiphysics was employed for modelling and optimization. The sensor demonstrates an exceptional sensitivity of 1000 GHzRIU⁻¹ and a detection limit of 0.310 RIU. In addition, the designed sensor maintains a high-quality factor of 8.176 and exhibits stable performance across its operational frequency range (0.1–1.4 THz). Moreover, the performance analysis under varying conditions showed consistent transmission patterns and frequency-dependent characteristics, with the graphene chemical potential that could be significantly influencing the sensor’s response. Furthermore, the integration of polynomial regression-based machine learning optimization yielded remarkably accurate predictions (R² > 0.97) across various operational parameters. Therefore, the investigated numerical findings and machine learning optimization prove that this sensor represents a significant advancement in tuberculosis detection technology, offering improved sensitivity and reliability compared to conventional methods, whilst maintaining fabrication feasibility through standard cleanroom processes.