Synopsis

We implemented B₁⁺ correction for our metabolite concentration estimation and analyzed the influence of B₁⁺ inhomogeneities. Since the B₁⁺ field is relatively homogeneous in the center of the brain anyways, the biggest effects were observed closer to the fringes of the brain. The concentration estimates increased between 1.0% and 5.6% for different metabolites when comparing B₁⁺ correction to simple T1 correction, remaining within the spectrum of values established by previous literature.

Introduction

One of the challenges faced by 7T magnetic resonance spectroscopic imaging (MRSI) is the inhomogeneity of the B₁⁺ field and the resulting data impairment. For this study, we implemented a B₁⁺ correction for our concentration estimate calculation¹ and quantified the effect of this B₁⁺ inhomogeneity correction. This abstract contains a preliminary evaluation.

Methods

We measured seven volunteers (4 male, 3 female, mean age 24.3±4 years) using a protocol consisting of an MP2RAGE, a CRT-FID-MRSI (64x64x39, FOV: 220×220×133 mm, 3.4 mm isotropic voxel size, TR=450 ms, acquisition delay 1.3 ms, 39° flip angle, TA=15 min, WET water suppression) and a water reference MRSI scan (TR=200 ms, 27° flip angle, TA=200 s), and a B₁⁺ map (128x128x10, FOV=250x250x113 mm, TR=6500 ms).
The spectroscopic data was post-processed as described in previous work^1,2 and subsequently spectrally fitted using LC Model³.
The B₁⁺ correction process consisted of multiple steps which have been compiled in figure 1. First, the measured B₁⁺ map was interpolated to the desired matrix size of 64x64x39. Since it was acquired using a sinc pulse and a different transmitter voltage U than the MRSI sequence, the B₁⁺ map had to be modified using a simulated MRSI slice profile P^MRSI with a full width at half maximum of 133mm, according to
$$B_1^{MRSI} = B_1^{Measured} P^{MRSI} U_{Used} / U_{Ref}$$
Next, for every voxel, the anticipated signal was calculated⁴ based on the local flip angle ɑ_L and the respective TR and T1 for each of the metabolites of interest (as well as water):
$$S_{B1} = M_0 sin(\alpha_L) (1-e^{-TR/T1})/(1-cos(\alpha_L) e^{-TR/T1})$$
For comparison, a signal map S_T1 containing purely T1 correction terms was created analogously, with the only difference being that the local flip angle ɑ_L was replaced by the Ernst angle ɑ_E, thereby neglecting any local B₁⁺ effects.
Due to the short acquisition delay used, T2 relaxation effects were omitted. The signal maps were calculated for grey matter (GM) and white matter (WM) and subsequently linearly combined based on FAST segmentation.
The amplitude maps of the metabolites and water, which generally contain values in institutional units, were normalized with regards to the influences of the B₁⁺ field via division by the corresponding signal maps, and in conjunction with literature water concentrations C_Water in GM and WM, internal water referencing was performed¹.
$$C_{Met} = C_{Water} A_{Met} / A_{Water}$$
The resulting 3D map C_Met describes the estimated concentration of a metabolite. For numerical analysis, averages and standard deviations of each concentration map were calculated for the entire brain as well as a single-slice volume of interest, both for T1 and B₁⁺ correction. We also analyzed the calculated signal maps in both cases.

Results

The signal maps S_T1 and S_B1 differed by 7.5% to 10% when B₁⁺ correction was performed. Together with the water signal map, this resulted in changes of the metabolite values by 3.8% for total choline (tCho), 1.0% for total creatine (tCr), 2.0% for glutamate (Glu), 5.6% for myo-inositol (Ins) and 2.9% for N-acetylaspartate (NAA). In a slice in the upper brain, where B₁⁺ inhomogeneities are relatively small, the increase was only around 3.0% for tCho, 0.7% for tCr, 1.4% for Glu, 4.5% for Ins and 2.3% for NAA, and the deviation of the signal maps was smaller as well (see table 1).
In absolute terms, the concentration estimate with T1 correction was 1.82±0.82 mM for tCho, 6.63±2.69 mM for tCr, 8.92±4.89 mM for Glu, 5.52±2.03 mM for Ins and 9.97±4.15 for NAA. B₁⁺ correction increased those values to 1.89±0.86 mM for tCho, 6.70±2.74 mM for tCr, 9.09±5.04 mM for Glu, 5.83±2.15 mM for Ins and 10.25±4.31 mM for NAA. This difference is relatively small and changes in the maps are mainly visible in border regions, where the B₁⁺ field becomes very weak, resulting in a significantly lower excitation and thus a bigger improvement due to the correction.
Figure 2 shows B₁⁺ maps, figure 3 displays concentration estimate maps of NAA together with the corresponding signal maps, and lastly, figure 4 shows concentration estimates for tCho, tCr, Glu, Ins, and NAA.

Conclusion

B₁⁺ correction has a small effect on metabolite concentration estimates for our method. Since B₁⁺ effects vary significantly within the FOV, the correction is more significant in some areas than others, and it can be valuable when the regional evaluation is desired. The B₁⁺ correction increased the concentration estimates compared to our previous publication, which had arrived at values on the lower end of the spectrum given by other literature¹.
This work is limited by the small number of volunteers and the resolution of the B₁⁺ map. Improvements in B₀ shimming as well as parallel transmit RF pulses could further improve our method. The development of a voxel- or ROI-wise analysis pipeline as a replacement for map averages should also lead to more reliable data.

Acknowledgements

This work was supported by the Austrian Science Fund (FWF) grants KLI-646, P 30701 and P 34198.