Phosphorus (³¹P) MRS of human brain has been conducted on 7T scanners to achieve better SNR and spectral resolution.^1-5 The technique of Nuclear Overhauser Effect (NOE) has been utilized to further increase SNR of ³¹P signals^1,4,6, where either WALTZ or stochastic waveform was continuously applied during relaxation time using low RF peak power. Although NOE using low RF peak power has been successful it is inherently narrow band. At higher field accompanied by larger chemical shift dispersion of the protons, broadband saturation of protons is desirable. Broadband NOE can be implemented using a train of evenly spaced short hard pulses with a small duty cycle (<~1%). This approach allows high RF power to be used on short pulses, therefore, provides wider bandwidth. It has been successfully applied previously to ¹³C MRS of human brain at 7T⁷. This study is to test the feasibility and effectiveness of broadband NOE for ³¹P MRS of human brain.

Hardware: A Siemens 7T scanner with VB17 software was used. Proton coil was a shielded quadrature half-volume coil (two overlapping octagon loops, nominal size=12.7x12.7cm²). ³¹P coil was a surface coil (diameter=7.0cm), as indicated by the red line in Fig. 1. The coils were connected to the scanner via an interface box (Quality Electrodynamics) containing T/R switches, preamplifiers and filters for both channels.

³¹P MRS: A 6x6x6cm³ voxel in the occipital lobe (white box in Fig. 1) was shimmed using all 1^st and 2^nd order shims and five 3^rd order shims. The typical water linewidth from the voxel was 15~17Hz. ³¹P spectra were acquired using an excite-acquire sequence: hard pulse=250ms, TR=3s, number of data point=1024, and SW=5kHz. ³¹P flip angle was empirically optimized to obtain maximum SNR for a given TR.

NOE pulses: A train of 26 equally spaced hard pulses (pulse width=0.5ms, nominal flip angle=180^o) were applied in proton channel during relaxation time for NOE. For TR=3s, the duty cycle of NOE pulse train was 0.43%. From the applied transmit voltage reported on Siemens console, the average power for NOE was 3.5W. The actual power delivered into the RF coil would be 25% less due to the additional power loss in the system.

Data process: The FID was left-shifted by two data points to reduce the broad hump in baseline from immobile phosphate signals. The first 512 data points in FID were used and zero-filled to 32k and broadened with LB=1Hz. No additional baseline correction was applied.

An axial gradient-echo image (Fig. 1) demonstrates that the proton coil offers adequate B₁ field homogeneity for NOE and decoupling within the effective volume of the ³¹P coil. Fig. 2 shows a typical in vivo spectra from a heathy volunteer with TR=3s and NA=128 without NOE pulses. The spectra show that eleven phosphorus resonances were distinguishably detected including: phosphoethanolamine (PE), phosphocholine (PC), inorganic phosphate (Pi), glycerophosphoethanolamine (GPE), glycerophophocholine (GPC), phosphocreatine (PCr), γ-, α-, β-adenosine triphosphate (ATP), nicotinamide (NAD), and uridine diphospho glucose (UDP). Fig. 3 shows the spectra from a subject (TR=3s, NA=64) without NOE (a) and with NOE pulses (b). The average percentage increases of major resonances from three subjects due to NOE are listed in Table 1.As shown in Fig. 2, resonances of PE and PC and that of γ-ATP and NAD are well resolved at 7T. This excellent spectral resolution is contributed by larger frequency separation at 7T and by applying 3^rd order shims in the occipital lobe. The results in Table 1 are in broad agreement with literature results using WALTZ¹. A detailed comparison with NOE by WALTZ will be made in the near future. Because of the very low duty cycle of NOE hard pulses, much stronger RF pulses can be used to generate broadband NOE safely.