Introduction

MR image quality can vary across sites due to differences in experimental conditions, sequence implementation, and image reconstruction and preprocessing choices. Significant progress has been made to making data acquisition and post processing more consistent across sites and vendor platform [1-2]. Another aspect that directly affects the image quality in functional, diffusion and perfusion MRI as well as magnetic resonance spectroscopy (MRS) is B₀ inhomogeneity. In fact one of the most important challenges to obtain good MRS data quality is the inhomogeneity of the static magnetic B₀ field in different regions of the human brain and body [3]. Poor B₀ inhomogeneity renders adjacent spectral peaks separated by only few hertz indistinguishable and reduces the accuracy and precision of metabolite concentration estimates. In this work, we have developed a versatile open source B₀ shimtool that combines several recent algorithmic developments [3-8] and considers recommendations of a recent consensus paper [3]. We have compared the B₀ shimming performance of the open source B₀ shimtool against the vendor implemented autoshim algorithm for different shim problems (i.e. single voxel, multi voxel, ROI) and evaluated the resulting spectroscopy data quality.

Methods

Software tool: This open source B₀ shimtool is generally implemented in MATLAB (Natick, MA, USA). In addition, we have incorporated FSL packages like ‘prelude’ and ‘bet’ for phase unwrapping and Brain skull stripping respectively. Real shim fields of all shim coils are considered by decomposition of respective acquired B₀ calibration maps into a linear combination of spherical harmonic functions yielding respective spherical harmonic coefficients for each shim channel [6]. Since previous work reported that an iterative pseudo inverse method to solve the subject specific shim problem shows the most robust convergence for most B₀ shimming problems the B₀ shimtool returns optimized shim values using the respective ConsTru (constrained TSVD inversion method) algorithm [3]. This approach considers maximum available shim field strength by truncating the smallest singular value of spherical harmonic function and reinvert the new matrix in an iterative manner unless all the shim values fit inside our hardware constrains [5]. With the shimtool we can import square or rectangular spectroscopy voxels, draw random shim ROI shapes in any specific region or we can select whole brain shimming based on automatic skull stripping. Also we have implemented the region of less interest (ROLI) approach to consider shimming needs outside of the shim ROI for better shimming and artefact control in some regions [5]. Finally, automatized calculations of slice specific optimal shim values as input for dynamic shim experiments [7]. In principle the B₀ shimtool is vendor and field strength independent.
Experimental validation: We evaluated the shim tool at a 3T MRI scanner equipped with 2^nd order shim coils (Siemens Prisma). First, we obtained real shim field calibration data of the B₀ fields produced by gradient coils and second order shim coils using a 3D GRE sequence. This calibration process needs to be performed once and it is unique for each system. Subsequently, we performed different phantom and in vivo spectroscopy experiments in four subjects after IRB approval and signed consent to test the B₀ shimtool performance: (i) 2^nd order single voxel shimming in the frontal and occipital cortex of the human brain in vivo and (II) 2^nd order shimming for 1H MRSI in the human brain using a global brain shim ROI versus a ROI matching the MRSI FOV. To evaluate respective B₀ shimming performance single voxel ¹H MRS data were acquired using a PRESS sequence (FOV = 20x20x20 mm³, TE = 30 ms, TR = 2sec, Angulation = 0) and ¹H MRSI data were recorded using a semi laser sequence chemical shift imaging (CSI) sequence (FOV = 95x125x20 mm³, TE = 40 ms, TR = 1.5sec, Angulation = 0) respectively. We cross-validated our B0 shimtool against the vendor implemented default shim routine We have compared the metabolite spectrum and metabolite maps we are getting from different brain regions for the vender’s implemented algorithm and our shimtool.

Results & Discussion:

Figure 1 shows the user interface of the B₀ shimtool and Figure 2 the tools B₀ shim analysis capabilities. Figure 3 A and B shows the shimming performance of the B0 shim tool versus the vendors shim method in single voxel spectroscopy. Figure 4 A and B shows a comparison of 1H MRSI spectra after B₀ shimming with the vendors implemented shim method versus our B0 shimtool, while in Figure 5 the comparison of respective NAA maps are shown .
These results show that in single voxel spectroscopy applications the vendors shimming method shows similar performance as the B₀ shimtool in the OCC (easy to shim), while the B0 shimtool clearly outperforms the vendors shimming method in the frontal cortex. Similarly the B₀ shimtool leads to improved spectral quality in 1H MRSI. NAA maps look more homogenous if we do shimming using our custom made B₀ shimtool in comparison to the vendor implemented method.

Conclusion

We have introduced a versatile open source GUI based B₀ shimtool that provides better B₀ shimming combining real shim field calibration with the ConsTru algorithm for different shimming problems compared to vendor’s pre implemented shim in different human brain areas which are difficult to shim.

Acknowledgements

This work was performed at Advance Imaging Research Center At University OF Texas Southwestern Medical Center Dallas. This work was supported by Cancer Prevention and Research Institute of Texas (CPRIT) grant / RR180056.