A Machine Learning–Assisted Double-Negative Metamaterial Absorber for Prostate Cancer Detection

This paper presents a double-negative metamaterial absorber (MMA) based on a cross-connected square-ring resonator, which combines a square ring with four orthogonal rectangular arms. The proposed symmetric design has significantly improved field localization and enables dual-band absorption at 1.4 and 4.1 THz having absorption of 99.67% and 99.93%, respectively, making it suitable for polarization insensitive biosensing. The absorber is built on a polyimide substrate with silver layers on both sides, and the overall dimensions are 0.205λ × 0.205λ × 0.015λ. The proposed MMA shows strong sensitivity to refractive index changes, distinguishing between prostate cancer cells (n = 1.398) and normal prostate cell (n = 1.30). The calculated sensitivities are 2530 and 2875 GHz/RIU at the two frequencies, with figure of merit (FOM) of 78.13 and 57.5 and Q-factor of 41.57 and 81.8, respectively. Parametric studies including H-field, surface current distributing, and substrate variations are conducted to understand the absorber’s behavior. Furthermore, an equivalent circuit is proposed, to confirm a good agreement with computer simulation technology (CST) simulation result. To improve the predictability, a machine learning (ML) framework has been employed considering five advanced algorithm models including three test scenarios. Among all the results, the extra-trees regressor (ETR) gave the most stable and precise predictions, attaining a validation accuracy of 99.78%.