Super perfect polarization-insensitive graphene disk terahertz absorber for breast cancer detection using deep learning

• Published in: Optics and laser technology 2025-05, Vol.183, Article 112246
• Main Authors: Zamzam, Pouria, Rezaei, Pejman, Khatami, Seyed Amin, Appasani, Bhargav
• Format: Article
• Language: English
• Subjects: Biomedical sensing; Breast cancer; Graphene; Metamaterial; Refractive index sensor; Terahertz absorber
• ISSN: 0030-3992
• DOI: 10.1016/j.optlastec.2024.112246

•The structure’s impedance is carefully matched, leading to maximum absorption at specific frequency.•THz frequencies, surpassing the absorption capabilities of earlier models.•Incorporating a deep learning to enhance breast cancer detection, increases the sensing accuracy by up to 99.8%. In this research, a new analytical approach to terahertz absorbers using graphene disks is introduced. Graphene disk is one of the ubiquitous patterns of graphene that can be seen in every paper. The notion here is that all the relationships between this graphene disk and the impedances of the structure are examined in this paper. It is concluded that if both the dielectric permittivity and the dielectric layer height are appropriately calculated, the structure input impedance totally depends on the graphene layer impedance. This concept has been calculated for two layers of graphene and we claim that this idea will be valid for any other graphene layers. It is finally determined that the structure reflection coefficient depends on the radius of the graphene disks. That is the optimum amount of the radius of the graphene disk layer and it will bring about 100% absorption in the structure. Hence, all the necessary equations to elucidate this concept are brought in this paper. The proposed graphene disk absorber has been used for breast cancer detection. Breast cancer cells and normal breast cells have different refractive indices of 1.399, and 1.385, respectively and then they result in different absorption spectra. A deep learning model comprising two hidden layers has been trained with 504 different absorption spectra, obtained by varying the thickness of the breast cell tissue, polarization angle, and incidence angle. The accuracy of detection was observed to be 99.3% with the training data, and 99.8% with the validation data, demonstrating that combining deep learning with graphene-based THz absorbers can be a key contribution to the biomedical field of research.