Effective suppression of maternal fat is critical for functional imaging (fMRI) of the fetal brain while using echo planar imaging (EPI). It is also necessary to limit acoustic noise and avoid maternal peripheral nerve stimulation, which both constrain gradient switching rate, leading to reduced bandwidth in the phase encoding direction, increasing sensitivity to magnetic susceptibility. Localized image-based shimming (IBS) can resolve these issues for the fetal brain, but can provoke high field variation in the surrounding maternal adipose regions leading to poor fat suppression. We aim to address this issue by using IBS of the fetal brain, with linear constraints across maternal fat regions, and optimization of saturation pulse frequency offset.Shimming methods, combined with fat suppression by spectrally selective saturation pulses with inversion recovery (SPIR), were initially tested on adult volunteers using the kidney as a surrogate for fetal brain. Second order shimming for the target region only, and second order shimming for target with additional linear constraint on field offsets in maternal subcutaneous fat were compared. In each case the pulse frequency offset was simulated to determine optimal parameters for fat suppression. Avoidance of signal suppression in the target region in EPI and degree of fat suppression were used as quality measures. Local IBS with linear constraints was found to be most effective, so this was optimised and tested on seven pregnant volunteers (gestational ages ranged 27 to 31 weeks). Written informed consent was obtained prior to imaging. Data was acquired on a 3T Philips Achieva scanner using an 32-channel cardiac coil using an EPI acquisition designed for fetal fMRI (resolution of 2.5×2.5×3.5 mm³ and field-of-view (FOV) 320×320 mm²; BW_pe=13 Hz/pixel) and SPIR fat suppression (pulse-length= 7 ms, BW= 679 Hz and B1=5.7 µT). B0-maps (ΔTE=2.3 ms, resolution 2.3×2.3×10 mm³ and FOV 350×350×100 mm³) were acquired in each subject and phase unwrapped as necessary.¹ The fetal brain was delineated with a circular region of interest (ROI) and maternal fat was segmented from a separate 3-point Dixon acquisition. The following were compared: i) Optimal IBS (O-IBS) which minimizes inhomogeneities within the brain ROI only using up to second order shims, optimised with a least-squares algorithm regularized by the sum of high shim parameters (λ=0.03); and ii) Constrained IBS (C-IBS) which adds linear constraints on the fat region [F_0fat-300;F_0fat+100]Hz, where F_0fat is the resonance frequency of the fat.² These limits were chosen according with the optimized SPIR pulse from non-pregnant adult tests. Optimisations employed the lsqlin function from Matlab R2012. Residual B₀-maps were also acquired for each shim case (data for C-IBS is missing for one subject).

Data from the non-pregnant volunteers and the simulations showed that the optimal SPIR frequency offset for O-IBS was approximately 220 Hz (Figure 1). Larger frequency offsets began to saturate the water signal. Figure 2 illustrates performance for the scanner’s standard second order local shimming method (PB), based on a FASTMAP³, versus O-IBS and O-IBS with optimal offset in a pregnant subject. The constraint applied through the regularization improves the result from PB to O-IBS. A slight residual fat artefact remains even with the new offset. Simulation predicts further improvement with C-IBS and optimal offset (not acquired in this subject). O-IBS and C-IBS achieved similar field homogeneity within the brain ROI (Table 1). The percentage of the fat region outside the constraints ([-350; 100] Hz) decreased from 15±11% with O-IBS to 8±4% with C-IBS. Figure 3 shows the progressive improvement achieved going from linear (global) auto-shim, to O-IBS, and to C-IBS with optimised offset. Linear shims leads to severe distortion in the brain and default SPIR frequency suffers poor fat suppression. O-IBS (with default SPIR) minimises deformation but still has sub-optimal fat suppression. C-IBS with tuned SPIR offset corrected both problems in this example. Overall, in terms of fat suppression efficiency with IBS methods: O-IBS with default SPIR was the poorest, improving with the optimized pulse offset, while C-IBS with this offset was more stable showing the same or better result as the previous combination in all subjects tested.

As figures 2 and 3 show, poor fat supression can cause contaminating signals to fold onto the fetal brain in EPI images. The results showed that is possible to obtain more complete fat supression when combining an optimized SPIR offset with a constrained shimming approach without compromising fetal brain shim. Further improvements could potentially be made with new SPIR pulse design and consideration of B₁ field uniformity.