Heart disease frequently involves impaired cardiac energy metabolism, which can be monitored non-invasively by phosphorous magnetic resonance spectroscopy (³¹P-MRS).¹ We have used a 16-channel ³¹P receive array (Rapid Biomedical) which gives excellent receive SNR but which (on an 8kW Siemens 7T scanner) has a peak B₁⁺ of only ~10µT at the depth of the heart, due to its 28x28cm² transmit element.² A more efficient transmit element would allow increased SNR for cardiac ³¹P-MRS by exciting at the Ernst angle; or even excitation with adiabatic pulses for absolute metabolite concentration measurements. This work aims to develop a ³¹P/¹H coil with optimised transmit for cardiac ³¹P-MRS at 7T.

Simulations: The following coil designs were simulated in CST Studio 2014: the 28x28cm² square transmit element in the 16-channel array²; 10cm, 15cm, 20cm and 30cm diameter circular loops; and 10cm, 15cm and 20cm diameter overlap-decoupled quadrature pairs.³ The “Laura” voxel model (CST) was used for loading, with FIT calculation using >1 million mesh cells, and co-simulation for the (balanced) matching networks. All S_ii were <-30dB at 120.3MHz. We aimed to maximise B₁⁺ across the heart, and also to minimise SAR10g / (B₁⁺)². To obtain safety parameters, we then simulated (Fig. 1) the 15cm ³¹P quadrature pair with an overlying 10cm ¹H loop at both 297.8MHz (¹H) and 120.3MHz (³¹P) frequencies on the Laura and Gustav virtual humans (each in two orientations) and on a large cube of muscle.

Coil: We fabricated the coil (Fig. 2) from 2mm copper wire, using Series 100E fixed capacitors (ATC Corp), and SGNMC1206E variable capacitors (Sprague Goodman), in a housing repurposed from a Siemens 1.5T cardiac coil. The two ³¹P loops were decoupled by adjusting their overlap. The ³¹P and ¹H loops were decoupled by adding inductor-capacitor-capacitor (LCC) traps⁴ to the ³¹P loops (L ~100 nH, C_s = 9.2pF, C_p = 3.3pF). The coil was connected by coaxial cables to a “T/R switch box” (Virtumed LLC, MN, USA) containing 3x preamps for the loops, and a 90° quad-hybrid. Balanced matching was sufficient to suppress common mode currents⁵, although foam sleeves (FG265, Component Force Ltd) were fitted around the cables as a further precaution.

Phantom tests: Losses from the LCC trap circuits were tested by comparing B₁⁺ and B₁^- maps made from a set of ³¹P FLASH images at 11 flip-angles on a 1L triphenylphosphite (TPP) bottle (Fisher Scientific UK Ltd, used as supplied). TPP has a 3.8M ³¹P concentration, 725ms T₁ and a low conductivity and permittivity. TPP is a liquid at room temperature, is not toxic to humans, and costs <£50 per litre.

The SAR simulations were validated by heating turkey mince in the MRI scanner (Siemens 7T) while monitoring with fibre optic temperature probes (Neoptix).^6,7

B₁⁺ was measured using a phantom⁷ comprising 14L saline for loading and a moveable 2x2x2cm³ cube of KH₂PO_4(aq) inside. At each depth, non-localised spectra were recorded at several transmit voltages and the peaks integrals were fitted to a sinusoid to determine B₁⁺.

In-vivo tests: Three healthy volunteers were scanned using our standard cardiac ³¹P-MRS protocol (3D UTE-CSI, 1s TR, 16x16x8 matrix over a 24x24x20 cm³ field of view) with the new coil in the heart, leg and liver. Localisers were acquired with single-shot FLASH and pulse-oximeter gated CINE FLASH. In the liver, B₀ shimming was performed using a dual-echo GRE method.⁸

Simulations: Table 1 and Fig. 1 show the simulation results. The coil with the greatest B₁⁺ and the lowest SAR/(B₁⁺)² over the heart was the overlapped 15cm quadrature pair (Fig. 2).

Phantom: Losses due to the LCC trap were <10%, consistent with previous findings.⁴ Defining 1=³¹P_A, 2=³¹P_B, 3=¹H, S-parameters on the 14L phantom in the magnet were S₁₁ -30dB, S₂₂ -28dB, S₁₂ -24dB at 120.3MHz, and S₃₃ -32dB, S₁₃ -32dB, S₂₃ -25dB at 297.8MHz. Similar values were obtained in vivo. In the 14L phantom, peak B₁⁺ varied between 16 – 25µT at the depth of the heart (Fig. 3).

In-vivo: Localization was straightforward and good quality spectra were recorded in the leg, liver and heart. Fig. 4 shows representative cardiac spectra and localizer images with acceptable SNR right across the septum. In another abstract, we show in 5 normal volunteers, that this coil enables adiabatic excitation across the heart for the first time at 7T.

This transmit coil delivers the B₁⁺ needed for cardiac ³¹P-MRS at 7T. Adding a ¹H loop makes localization and per-subject B₀ shimming straightforward. This transmit coil should now be combined with our receive array to optimise both B₁⁺ and B₁^-.