Design investigation of metasurface sensor with Square pixels and neural network optimization for early cancer detection

Abstract

In this article, a method for designing a THz metasurface sensor with the help of a neural network is presented. In this method, the goal is to obtain a metasurface with the highest Q-factor. For this purpose, the upper level of the metasurface is considered to be 9 pixels, and with the help of the backpropagation neural network, the optimal structure is obtained with the maximum Q-factor. The Q-factor obtained from the optimal structure is 1500 for 0.48 THz. Finally, this metasurface is used as a sensor for discriminating the cancerous tissue from healthy tissue for skin and breast. Results show that this structure can distinguish healthy and cancerous tissues based on the different effects of them on the Q-factor, the frequency of the first reflection peak, and the value of the first reflection peak based on full wave method of finite integrated technique (FIT).