High-sensitivity terahertz metasensor for cervical cancer Diagnosis: Graphene modulation and XGBoost-Assisted optimization

Abstract

Cervical cancer remains a major cause of mortality, particularly in low-resource settings where traditional cytology-based screening faces challenges such as limited infrastructure and trained personnel. To address this, we present a terahertz (THz) graphene-enhanced metasurface biosensor enabling rapid, label-free detection of cervical cancer biomarkers without complex sample preparation or expensive labs. Using finite element method (FEM) simulations, we demonstrate that tuning graphene’s chemical potential from 0.1 to 0.9 eV significantly modulates peak absorption from 0.223 to 1.316, providing a wide dynamic range for sensitive detection across varying sample concentrations. The sensor exhibits robust angular stability, with absorption increasing from 0.546 to 1.306 as the incident light angle shifts from 0° to 80°, ensuring reliable performance without precise optical alignment. Refractive index sensing experiments reveal frequency shifts of 50 GHz and consistently high absorption (55.16% to 56.54%), achieving a sensitivity of 300 GHz per refractive index unit (RIU) and a figure of merit of 12 RIU -1 . To enhance diagnostic accuracy, we integrated an XGBoost machine learning algorithm that analyzes the complex spectral data, achieving 86% prediction accuracy with low error rates. This combination of advanced sensing and AI-assisted analysis offers a promising, cost-effective solution for cervical cancer screening in resource-limited environments.