Data-Efficient Prediction of the Specific Absorption Rate in a Human Head Model Exposed to a Plane EM Wave Using Gaussian Process Regression

In this contribution, the specific absorption rate (SAR) in a human head model exposed to a 13.56 MHz external electromagnetic (EM) wave is predicted using Gaussian process regression (GPR). The electrical properties of the tissues are varied around the nominal values by $\pm 20 \%$ to account for material uncertainties. The GPR model achieves an $\mathbf{R}^{2}$ regression score higher than 0.99. The maximum and the worst-case exposure over the variation space are found using a tenth of the samples needed by random sampling on average. This allows the fast comparison against standards with a high degree of confidence. A sensitivity analysis of the exposure is extracted from the GPR model to provide an insight into the effects of the different parameter variations. The proposed framework is applicable to radiated susceptibility scenarios and beyond.