Preliminary studies have proved the feasibility of head MR imaging with a dipole array on the world’s first whole-body 10.5T (450MHz) MRI^1,2,3. But the dipole structure can take various forms. This work reports an extensive research effort to optimize the existing dipole array for head imaging through numeric modeling.

In order to optimize the dipole array structure for head imaging at 450 MHz, the following factors were taken into consideration: 1) array format: dipole arrays terminated with floating end-ring sections at both ends, dipole array terminated with end-ring sections only at the bottom while the upper half were replaced with straight wires or wires conformal to the head geometry, 2) dipole format: straight wires with meander structure or lumped L/C for frequency tuning, 3) array length: dipole arrays with lengths ranging from 160mm to 237.5mm were taken into consideration, 4) other design elements that may potentially elevate dipole array performance: RF shielding, global dielectric wrapping, dielectric/RF mirror (5mm thick x 240mm diameter) placed 7.5mm above head to limit power leakage, and 5) decoupling methods and their impact on the array performance. Some structures are illustrated in Figure 1.

The starting dipole array (Fig. 1-01) consists of eight 160mm long dipoles terminated with floating copper ring sections (20mm wide x 86mm length) on both ends. The dipoles are evenly distributed on a cylindrical surface (256mm diameter). If used, the RF shielding is 195mm (length) x 312mm (Diameter), Fig. 1-07. To assess dielectric structures (Fig. 1-08,09), the relative permittivity was set to 1, 5, 10, 20, 40 and 80. To evaluate decoupling, variable capacitors are placed among the bottom/upper end-ring sections in Fig. 1-11 to decouple neighboring dipoles.

All simulations were performed with the Finite Difference Time Domain method (XFDTD, Remcom, PA), loaded with a medium male size HUGO head model⁴. The electrical properties of its 17 tissues were adjusted for 450 MHz. All arrays were driven with quadrature-mode phasing.

All results were normalized to a total of 1 watt RF power dissipated within the head model. RF homogeneity was defined as the percentage of voxels whose |B₁⁺| was within 20% deviation from the averaged |B₁⁺| within the 3D FOV, as denoted by the green box in Fig. 1-08.

The |B₁⁺| distributions for selected array structures are presented in Fig. 2 for the orthogonal central brain slices. Figure 3 presents the current magnitude along coil length for three dipole formats (Fig 1. 04-06). Performance versus coil length is summarized in Figure 4 and Table 1.

The following observations are based on the coils being driven in quadrature and the specific loading position as illustrated in Fig. 1-08.

1) The starting design (Fig. 1-01) with quadrature drive generates strong center brightness, resulting in poor RF homogeneity. The addition of RF shielding exaggerates this pattern. Meanwhile, arrays with longer straight (Fig. 1-02) or conformal dipoles (Fig. 1-03) produce much more uniform |B₁⁺|. 2) Dipole arrays of same length have almost identical current distributions, regardless of the tuning elements (meander or lumped inductance or capacitance), as long as the meander size is small compared to the dipole dimensions or the gap among dipoles. 3) As the dipole length increases for structure Fig. 1-02, the mean |B₁⁺| in the FOV also increases, and the peak local 1gram and 10gram SAR decrease. (With one exception: when the dipole was 237.5mm long the peak 1gram SAR location shifted and defied the trend.) But the 210mm long array provided the most uniform |B₁⁺| in the FOV. 4) Global dielectric wrapping or a dielectric/RF mirror does not help elevate dipole array performance as applied. 5) Decoupling with capacitors on the end-ring sections effectively reduces the coupling between nearest elements from around -10dB to -20dB, but it also significantly lowers |B₁⁺| in the FOV. Other decoupling methods are being considered.

The next head dipole array should be 210mm long without upper end-ring sections, similar to Fig 1– 02.