Fast Algorithm for Noninvasive SAR Prediction Based on Artificial Neural Networks

In this communication, a fast algorithm based on artificial neural network (ANN) to noninvasively analyze and evaluate specific absorption rate (SAR) inside the equivalent head phantom is proposed. According to the fields outside of phantom, the inner fields of the phantom are calculated by field reconstruction. This process is accelerated by ANN. The network includes input/output layers and three essential hidden layers, representing fields and currents required for field reconstruction. The effectiveness of the proposed approach is validated by a case study of quick assessment of the SAR value within an equivalent human head phantom. The fast SAR prediction algorithm can control the error between the prediction result and the numerical calculation result within 4% in the range of 700 MHz–6 GHz, and at the same time reduce the solution time by more than 99%. Hence, the proposed algorithm can efficiently simplify the test process and reduce test time costs.