Outer Volume Suppression (OVS) applied to MR spectroscopy is a useful tool for improving localization, suppressing undesired signals, e.g. subcutaneous fat, and reducing chemical-shift displacement errors, which can result in signal reduction due to anomalous J-evolution. Localization is improved because of the very high bandwidth (>5kHz) and selectivity of saturation pulses while the selectivity of PRESS refocusing pulses is limited. At high B0 fields (≥3T) OVS can suffer from inhomogeneous B₁ and therefore can be improved by repeating OVS pulses multiple times [1, 2, 3]. The repetition time is however limited by SAR and timing constraints. Goal of the present work was to optimize OVS for inhomogeneous B₁ amplitude and for a broad range of T₁ relaxation times, and to apply it to PRESS in the brain at 3T.A sequence of 6 different OVS slices was repeated once, twice (2-train) or 3 times (3-train) and the flip angles were optimized using Bloch equation simulation and the fminsearch function in MATLAB. Minimization criterion was root mean square (RMS) of the remaining M_z magnetization at time of the PRESS excitation pulse within a range of ±10% B₁ amplitudes and 200-800ms T₁, while assuming T₂=T₁/10. Each OVS module (RF pulse and crusher gradient) had a duration of 5ms. A library of broadband Shinnar-Le Roux (SLR) pulses with polynomial phase response [4] and 6.5kHz bandwidth was designed for flip angles 75°, 80°, … 170°. PRESS was applied with CHESS water suppression and S-BREBOP pulses (broadband and B₁-robust refocusing with 2.8kHz bandwidth) [5]. Experiments were performed on MR750w 3.0T (GE Healthcare).

Optimal flip angles for the OVS slice directly before the excitation pulse were 90.8° for single, 85.0°, 102.1° for 2-train, and 76.5°, 96.5°, 109.7° for 3-train (Fig 1(a-c)). The optimization was repeated for different delay times between the last OVS module and the excitation pulse. E.g. for 31ms delay optimized flip angles were 95.1° for single, 79.4°, 115.9° for 2-train, and 96.7°, 85.4°, 172.5° for 3-train (Fig. 1(e-f)). Overall OVS performance was calculated by RMS M_z (Fig. 1(d)).

Localization performance was evaluated by acquiring an image of PRESS voxel in a homogeneous phantom containing vegetable oil. Line plots through the edge of the PRESS voxel with different OVS techniques demonstrate suppression performance (Fig. 2). OVS was tested in healthy human brain with a PRESS voxel close to the scalp (Fig. 3).

Compared to single OVS, the RMS M_z was reduced by 83% for 2-train and 95% for 3-train (Fig. 1(d)). For long delays B₁-robustness can be maintained but the dominating effect is T₁ relaxation, reducing saturation performance over the desired range of T₁ (Fig. 1(e-f)).

A single broadband SLR pulse did not sufficiently reduce signal outside the PRESS voxel, while the optimized 2- and 3-train OVS achieved good localization. The lipid contamination from subcutaneous fat at around 1.5ppm for the in vivo spectra was successfully removed with the 2-train OVS pulses (Fig. 3(d)). The presented 2-train OVS technique achieved high suppression despite B₁ inhomogeneity while reducing T₁ effects under the given SAR and timing constraints.