Synopsis

Keywords: Shims, Shims

B₀ shimming for single voxel spectroscopy (SVS) is of importance due to its role in determining linewidth, accuracy of the analysis and interpretation. However, for difficult brain regions, such as the prefrontal cortex, where shifts of 3-6mm in voxel position can dramatically alter the achieved linewidth in SVS, the lack of a reliable method to predict the achievable homogeneity results in failed studies and/or increased study times. In this work we demonstrate an objective and non-iterative method (SUMO) for voxel positioning that predicts and ensures adequate B₀ homogeneity is achievable.

Introduction:

Introduction: B₀ shimming for single voxel spectroscopy (SVS) is of importance due to its role in determining linewidth, accuracy of the analysis and interpretation. With relatively small volumes, ~8cc, B₀ variation is expected to be small and shimming with 2nd order spherical harmonics is generally thought to be sufficient. However, this assumption may not be valid in the anterior and inferior portions of the frontal and temporal lobes and some regions adjacent to the skull (Fig. 1). Methods to map the field in such regions with rapidly spatially varying B₀ fields have also been thought to be problematic due to limited accuracy of many 3D B₀ mapping methods, limiting the predictability of shimming. On a practical basis in difficult neocortical regions without clear landmarks and the probabilistic methods used in determining neocortical regions of interest (such as the prefrontal cortex, PFC), repositioning the voxel by a few mm can significantly improve B₀ homogeneity. With the alternative being discarding poor spectral quality data altogether, shifting the voxel position is the best option although costly insofar as the imaging time needed for the repositioning, shimming and reacquiring the data. However, we hypothesize that after collection of a single B₀ map of sufficient accuracy spanning the entire brain, the achievable B0 homogeneity for a voxel of arbitrary size, shape and location can be predicted and will ensure that adequate homogeneity is achieved.
The goals of this work were to: 1) accurately predict and achieve higher order shimming (1st-4th order) for SVS studies; 2) objectively and efficiently position the voxel to achieve a target B₀ homogeneity and 3) validate these methods in the PFC using LC modeling¹ of SVS measurements at 7T in the human brain.

Methods:

Data was acquired from 6 healthy subjects using a Siemens 7T Terra with a 16-channel transceiver array, and a 3rd and 4th order shim insert from Resonance Research Inc.^2-4. For anatomical visualization and positioning, a 1mm isotropic MP2RAGE image was acquired. Whole brain B₀ field maps at 3mm isotropic resolution (216x216x123mm³ FOV) were acquired using BOLERO⁵. This B₀ field map enables prediction of the achievable field homogeneity after localized shimming for arbitrary voxel sizes and positions over the entire brain. For the PFC, where small displacements on the order of 3-6mm can have significant effects on the achievable linewidth after shimming, the voxel position was optimized using Susceptibility Managed Optimization, SUMO. This process selects the voxel location by 1) manually selecting an initial target location (Fig. 2 red box), 2) defining a grid of potential voxel locations about the target location (Fig. 2 yellow box), 3) calculating the achievable homogeneity over that grid and 4) automatically selecting the nearest location to the target that achieves the user-determined homogeneity threshold (Fig. 2, green box). In this work a threshold of 5Hz SD was used. The location was shimmed (1st-4th order), SVS data acquired, and verification B₀ maps were acquired to document the achieved homogeneity. SVS data (20x20x20mm³) was acquired using a STEAM sequence⁶ (TR/TE/TM/NA of 6sec/8ms/20ms/32 and analyzed with LC Model¹.

Results:

Displayed in Fig.1 are color-coded B₀ (3mm isotropic resolution) images showing the predicted SD after 1st-2nd and 1st-4th order shimming (voxel 20x20x20mm³). As described, small displacements of a single or several B₀ map pixels (3mm) can decrease the linewidth by 50% or more (~10Hz reduction in FWHM). To demonstrate this effect, we evaluated the PFC with a 3x3 search grid in the AP and HF directions (Fig. 2). The RL position was fixed so that the voxel remained along the mid-line. Across 6 subjects, the predicted SD from the initial target location was 8.9+3.8Hz using 1st-4th order shimming. Application of SUMO using a threshold of 5Hz SD for 1st-4th order shimming, identified at least one acceptable voxel of <5Hz SD in 5 of 6 subjects (in the remaining subject the minimum SD within the search grid was 5.3Hz over the target ROI, which was used). The average distance between the initial target and the closest voxel meeting the threshold SD was 4.2±2.3mm. The measured SD after 1st-4th order shimming was 4.0+0.9Hz in excellent agreement with the predicted SD, 4.5+0.9Hz, calculated prior to shimming, demonstrating the accuracy of the B₀ maps. The predicted SD after 1st-2nd order shimming from this location was 7.3+2.5Hz, indicating the presence of inhomogeneities >2nd order in the SUMO selected voxel. LC Model analysis of the acquired spectra yielded a mean FWHM linewidth of 10.7+1.5Hz. Assuming a gaussian line-shape, this is equivalent to an SD of 4.6+0.6Hz (Fig. 3). The predicted and achieved linewidths for 1st-4th order shimming calculated from the B₀ maps, assuming gaussian line shapes, are 10.5+2.1 and 9.5+2.1 respectively, which are in excellent agreement with the LC Model calculated linewidth from the spectra, 10.7+1.5Hz.

Discussion/Conclusions:

We have demonstrated that small spatial displacements (3-6mm) in difficult brain regions such as the PFC can have dramatic effects on shimming. BOLERO B0 maps can accurately predict the achievable homogeneity which is in good agreement the measured linewidths in the acquired spectra. Thus, SUMO can provide an objective and non-iterative method for voxel positioning to ensure spectral quality.

Acknowledgements

NIH R01 – EB024408