High-Sensitivity Graphene-Based Terahertz Biosensor for Cervical Cancer Diagnosis: SVM-Assisted Optimization

This work introduces a high-sensitivity graphene-based terahertz biosensor optimized using a support vector machine (SVM) approach for accurate cervical cancer diagnosis. The proposed structure demonstrates strong reconfigurability, with the resonance frequency shifting from 4.84 THz at$\mu _{c} = 0~\text {eV}$to 5.03 THz at$\mu _{c} = 0.5~\text {eV}$, confirming the efficient tunability enabled by graphene’s chemical potential. Sensitivity analysis reveals distinct responses for representative biomarkers, yielding 57.6, 76.9, 100.3, and 116.9 (GHz/RIU), respectively. To enhance predictive reliability, a SVM regression model was implemented, achieving an excellent coefficient of determination of$R^{2} =0.978$. After optimization, the predicted sensitivities improved to 93, 129.2, 171.4, and 231.6 (GHz/RIU), demonstrating the model’s capacity to accurately capture nonlinear dependencies between chemical potential, temperature, and relaxation time. These results confirm that modulation of graphene’s electronic properties plays a decisive role in resonance control and sensitivity enhancement, establishing a compact, label free, and machine-learning-assisted platform for early detection of cervical cancer.