Coil designs motivated by the ideal current patterns corresponding to the Ultimate Intrinsic SNR (UISNR) have been used to boost central SNR at 3T and 7T [1-5]. For a cylindrical phantom and a current distribution defined on a concentric cylindrical surface, the ideal current pattern for optimal central SNR includes both divergence-free and curl-free components [6, 7]. At low field, divergence-free current patterns saturate the UISNR and arrays with an increasing number of loops can approach the UISNR [8-10]. While loops are exclusively divergence-free, recent work has shown that electric dipole antennae include both divergence-free and curl-free current components [11,12]. Here we explore in simulation whether arrays with an increasing number of electric dipole antennas can approach UISNR in the center of a head-sized phantom at 7T, and investigate selected practical design considerations.

FDTD simulations were performed with Microwave studio (CST, Darmstadt, Germany). Dipole arrays with various numbers of elements were simulated using 4 mm diameter rods, placed 10 mm away from a 20 cm diameter cylindrical phantom (ε_r = 52.5, σ = 0.5642 S/m) [Fig 1(a-e)]. All dipole arrays were limited to a 20 cm area along the z direction to ensure a fair comparison between arrays. In the two-row 32-element design, each element was shortened to 16 cm and nearest neighbors were offset along z, to reduce coupling and maintain the 20 cm range [Fig 1(d)]. Similar strategies were applied to the 24- and 48-element arrays with three rows, where each dipole was shortened to 9cm [Fig 1(c, e)]. All elements were tuned to 297.2 MHz with inductors close to the feed points.

For comparison, a three-row 48-element loop array was simulated for an identical phantom [Fig 1(f)]. The width and length of each loop element was adjusted to achieve geometric decoupling with neighboring coils. The length was also adjusted to fit three rows into the 20 cm range along the z direction. Three capacitors were distributed evenly along each element. All elements were tuned to 297.2 MHz.

In practice, the use of inductors to shorten dipoles introduces extra resistance into the array, particularly for very short dipoles, which may result in reduced SNR. To model this effect, we repeated the simulations including a model for the extra resistance (R) of the inductors. We calculated R on the bench using the Q value measured when the inductor was combined with a capacitor into a resonant LC circuit, such that $$$ R=2 \pi f L/Q $$$.

UISNR values were calculated using a full-wave electrodynamic simulation tool based on a current mode expansion and dyadic Green’s functions [6]. Separate UISNR optimizations were repeated including only curl-free, only divergence-free, and all current modes.

Figure 2 shows optimally combined SNR (i.e., matched filter combination including noise covariance matrix) for a transverse slice through the center of the phantom. The 48-element dipole array resulted in the highest central SNR among the six designs, with 28% higher SNR at the center than the 48-element loop array. All dipole arrays were found to exceed the UISNR for either divergence-free components or curl-free components at the center of the phantom, and the 48-element dipole array achieved 95% of UISNR for all currents (Figure 3). The 32-element array is the only 2-row design studied, which may account for its low SNR. When inductor losses were included in the simulations (Figure 4), the performance of the 48-element dipole array dropped to 87% of UISNR for all currents. When SNR was reconstructed using a sum-of-squares combination of the array elements without incorporation of the noise covariance matrix, due to lower inter-element coupling the 24-element dipole array outperformed the 48-element design (Figure 5), achieving 82% of the UISNR for all currents, and still exceeding the UISNR for divergence-free currents alone.Recent work has shown that the current pattern in an electric dipole antenna includes both curl-free and divergence-free components. This is supported by our simulation results showing that the SNR of dipole arrays can surpass the UISNR for either curl-free or divergence-free currents. We showed that, theoretically, a dense dipole array with the proper configuration could closely approach the central UISNR for all-currents, outperforming considerably any conceivable loop array for head imaging at 7T. However, in practice, inductor losses and inter-element coupling may limit gains with high element counts. Future work will include constructing a dense array of dipole antennas to confirm our predictions that loops are not needed for 7T head imaging.