Design and Optimization of a Graphene-Enhanced Terahertz Metasurfaces Surface Plasmon Resonance Biosensor with Multi-Material Architecture for Cancer Detection Integrating 1D-CNN Machine Learning for Performance Prediction and Analysis

This study presents the design, simulation, and analysis of a terahertz metasurfaces biosensor for cancer detection, integrating graphene, gold, silver, and copper in a multi-resonator configuration. The sensor architecture comprises a copper rectangular ring resonator, two identical gold-silver square resonators, and a graphene-coated circular resonator on a silicon substrate. COMSOL Multiphysics simulations demonstrate the sensor achieves a sensitivity of 1000 GHzRIU(-1) within the refractive index range of 1.36-1.401 RIU, with a figure of merit of 23.256 RIU-1 and quality factor ranging from 6.023 to 6.279. The sensor exhibits consistent performance with a detection accuracy of 23.256 RIU-1 and measurement uncertainty of 0.001. Additionally, the device demonstrates potential as a 2-bit encoder through graphene chemical potential modulation, producing distinct transmittance patterns for binary states. One-dimensional Convolutional Neural Networks (1D-CNNs) optimized and predicted sensor performance, achieving R-2 values between 97 and 100% in various parametric analyses. The proposed sensor's high sensitivity, coupled with machine learning optimization, presents a promising platform for early-stage cancer detection and biosensing applications.