Deep-learning-assisted terahertz biosensing using MXene–graphene metastructures for male fertility evaluation

This study presents an MXene–graphene biosensor architecture for high-sensitivity spermatozoa detection operating in the terahertz spectrum (0.1–1.0 THz). To receive the best performance of the designed sensor, we have investigated a wide optimization process for the different governing parameters, including the chemical potential of graphene, in addition to the thicknesses and dimensions of the considered materials. Meanwhile, the optimization studies reveal that increasing graphene chemical potential reduces optical transmittance, while larger resonator dimensions enhance absorption and induce bathochromic shifts. In contrast, the device achieves exceptional performance metrics: sensitivity of 10 000 GHz RIU−1, figure of merit of 123.457 RIU−1, and detection limit of 0.014 RIU based on the optimum values of the different parameters. Moreover, the statistical analysis establishes an inverse correlation between sperm concentration and resonance frequency (R2 = 0.70). A 1D-CNN regression model achieves 91% accuracy across refractive indices and 96% across incident angles. This platform represents a significant advancement in male fertility diagnostics, offering superior sensitivity compared to existing technologies.