Computation of electromagnetic fields for high-frequency magnetic resonance imaging applications.

A numerical method is presented to compute electromagnetic fields inside a 2 mm high resolution, anatomically detailed model of a human head for high-frequency magnetic resonance imaging (MRI) applications. The method uses the biconjugate gradient algorithm in combination with the fast Fourier transform to solve a matrix equation resulting from the discretization of an integrodifferential equation representing the original physical problem. Given the current distribution in an MRI coil, the method can compute both the electric field (thus the specific energy absorption rate (SAR)) and the magnetic field, also known as the B1 field. Results for the SAR and B1 field distribution, excited by a linear and a quadrature birdcage coil, are calculated and presented at 64 MHz, 128 MHz and 256 MHz, corresponding to the operating frequencies of the 1.5 T, 3 T and 6 T MRI systems. It is shown that compared with that at 64 MHz, the SAR at 128 MHz is increased by a factor over 5 and the SAR at 256 MHz is increased by a factor over 10, assuming the same current strength in the coil. Furthermore, compared with the linear excitation, the average SAR for the quadrature excitation is reduced by a factor over 2 and the maximum SAR is reduced by a factor over 3. It is also shown that the B1 field at high frequencies exhibits a strong inhomogeneity, which is attributed to dielectric resonance.