2D EPI-based pulses designed for the selection of limited profiles offer numerous advantages for zoomed DWI such as minimization of geometric distortions [1-4]. Examples of limited profiles in the abdomen are shown in Figure 1A. Despite of the demonstrated potential [2,3], incomplete fat suppression is often observed in zoomed DWI, which limits its applications in clinical practice [1,3,4]. Due to a sophisticated interplay between field inhomogeneities and chemical shift, the residual fat signal is distorted and might overlap with the main ROI in the images (Figures 1B, C).

B₀ shimming methods with constraints were demonstrated previously to restrict the fat frequencies [5]. Therefore, if an additional constraint is included such that the fat frequency during a 2D pulse is shifted outside the main ROI, the fat fraction will not be excited. In this experimental situation, the fat signal can be effectively suppressed after application of the RF pulse.

In this study, we simulated volume-selective B₀ shimming with constraints for fat. Our main goal was to explore the feasibility of shimming in both the main excitation ROI and the fat region, in order to accomplish an intrinsic fat suppression during 2D EPI-based pulses

Experiments were conducted on a 3T MRI system (Siemens Magnetom Trio). Sagittal multi-slice double echo GRE scans were acquired in the abdomen for a FOV of 32×32 cm², matrix of 128×128 and ΔTE of 2.46 ms. Field maps were reconstructed and used in simulations for shimming. Masks representing limited profiles and fat were defined. Two shimming approaches were investigated. Firstly, a constraint was tested such that the fat frequencies were limited to a multiple of 100 Hz, to explore the residual inhomogeneity in the main ROI. In the second simulation, the constraint was placed on the fat shifts during the 2D pulse caused by chemical shift. In this model, the shifts for fat were restricted to spatial area defined by aliased locations of the main excitation ROI. Several restriction areas for anterior and posterior fat fractions were investigated independently, in order to find an optimal spatial area leading to a minimal residual inhomogeneity in the main ROI. The shimming condition was that the shifted anterior and posterior fat fractions did not overlap with the aliased replicates of the main profile, corresponding to an intrinsic fat suppression during excitation. In addition, resulting field maps were simulated and used to predict distortions of the fat and main ROI [6].

Finally, the excitation of limited profiles was tested in a phantom with oil shown in Figure 4A. The selected profiles were scanned using a GRE sequence for shim currents varying along the PE direction of the designed 2D pulses, in order to detect a decrease in the fat signal.

Simulations for shimming with constraints showed that the fat frequencies could be restricted to a range on the order of 200 Hz for an experimentally suitable homogeneity in the main ROI (Figure 2). Figure 3A shows residual inhomogeneity in the excitation ROI when the constraints for shifts of the anterior and posterior fat were varied independently during the optimization. The topology of the resulting curve suggests that there are multiple solutions minimizing the inhomogeneity. As shown in Figures 3B and D, the best shimming scenario restricted the fat fractions to be shifted to the regions close to their original locations but the main ROI appeared to be distorted in the simulation. When fat fractions were constrained to more distant aliased regions, the main ROI was distorted more significantly and shifted outside the body (Figures 3C and E), which would mean a total signal loss in an actual imaging experiment. Scanning of the phantom with an oil shell showed that the fat signal was suppressed partially after 2D excitation when the shim settings were modified (Figure 4B-D). These effects are indicated by arrows in the images.

B₀ shimming with constraints introduces new degrees of freedom, allowing for intrinsic fat suppression during 2D excitation in zoomed DWI applications. However, the residual inhomogeneity in the main excitation ROI is inferior to the the standard shim procedures without constraints. As a next step, the influence of the frequency broadening of lipid resonances on the simulations presented above will be investigated.