31P-MRS provides unique information on brain energy and membrane metabolism in vivo. Alterations in brain high-energy and membrane phosphate metabolism were reported in subjects with a number of brain disorders including major depression, schizophrenia, substance abuse and tumors¹. The major limitation of brain ³¹P-MRS is low signal sensitivity at clinical scanners and from that resulting long acquisition times. Nowadays, several-fold higher SNR is available by using ultra-high field human whole-body MR scanners. Additional SNR may be gained by the use of receive arrays and advanced coil combination methods (e.g. whitened singular value decomposition – WSVD)³. This was recently demonstrated at 7T with a 7-channel receive array covering mostly the occipital lobe and WSVD coil combination². The purpose of this study was to quantitatively compare the SNR of brain ³¹P-MRS coils capable of covering the whole brain at various B₀ field strenghts. SNR was compared using phantoms and in vivo in clinically acceptable measurement times at 3T (birdcage), 7T (23ch array) and 9.4T (27ch array).

Data were acquired on a 3T, 7T and 9.4T MR systems (Siemens Healthcare, Erlangen) using double-tuned coils (¹H/³¹P): at 3T a birdcage coil (RAPID Biomedical GmbH, Rimpar, Germany), at 7T a 23-channel receive array (QED-Quality Electrodynamics, Mayfield, OH, USA) and at 9.4T a 27-channel receive array (Max Planck Institute,Tuebingen, Germany) with the same design as in⁴. A point source 1cm³ cube (1M KH₂PO₄) submerged in a 2l cylinder filled with tap water (Fig.1A) and a 2l spherical phantom (10mM KH₂PO₄) were scanned at all field strengths. Fully relaxed pulse-acquire FIDs were acquired with variable pulse amplitude. Pulse amplitude with maximal signal was used for SNR analysis. Both, identical spectral widths (SW=7kHz at each fieldstrenght) to exclude the confounding effect of SW on the overall SNR (inversely proportional to square root of SW) and identical chemical shift ranges (in ppm unit, e.g. 2.2kHz,5.2kHz,7kHz for 3T, 7T and 9.4T) for more practical comparision of SNR were applied.Additionally a homogeneous spherical phantom was scanned with a fully relaxed (TR 30s) 12x12x12 MRSI sequence with 45º and 90º to allow local SNR comparison and dual flip angle B₁ mapping.

In vivo measurements were performed on three healthy volunteers (age 32.6±3.2 years). MRSI scans at all field strengths were performed within two weeks with the following parameters: 8x8x8 matrix, FOV 20x20x20cm³, TR=1.5s, NA=16, weighted acquisition, block pulse 500us, Ernst FA=45,50,53º acording to published T₁ times⁵ for optimal SNR/t^1/2 at each field strength, TA=11min, identical chemical shift ranges in ppm unit (e.g. SW=2.2kHz,5.2kHz,7kHz for 3T, 7T and 9.4T). The acquired SNR was defined as the ratio between the amplitude of first FID point (AMARES fit) and the standard deviation (SD) of the noise of last 200 points in the unfiltered FID. For the in vivo data 4 voxels from each volunteer and each scan were quantified (Fig.3 B – green box).

Fig.1. shows results of the pulse-acquire experiments with point source ³¹P signal. Data shows approximately 2-fold higher SNR at 7T than at 3T. 9.4T provided an additional -more than linear- increase. WSVD coil combination outperformed Brown’s coil combination. Especially with the 7T coil the necessity of advanced coil combination algorithms is apparent. Pulse-acquire experiments with the spherical phantom show only approximately linear increase of SNR, which reflect the degraded B₁ fields at higher field strengths and thus an inhomogenous 90° excitation. A representative transverse slices of the 3D ³¹P-MRSI data from one volunteer are displayed in Fig.2. Data are displayed after noise normalization and equivalent time domain filtering. Fig.3A displays representative spectra of one volunteer and shows average PCr SNR from all three volunteers (Fig.3D). Noticeable is also the improved spectral resolution in the PME-PDE (2-7ppm) and α-ATP, NADP, UDPG (-7- -9ppm) regions. The results show experimental evidence of a more than linear increase in SNR with field strength in human brain ³¹P-MRS. Although we analyzed the central region of the RF coils (i.e., the region that benefits least from using receive arrays), our data show the dramatic increase in spectral quality at ultra-high magnetic fields. Increased SNR may be used for decreasing scan time and/or for increasing spatial resolution of ³¹P-MRS. Data are in good agreement with previous studies^5,6. However, the degraded B₁ homogeneity pose new challenges, which may be overcome with adiabatic RF pulses as commonly used for ³¹P-MRS for surface coils.