4019

Optimized B1/B0 correction for whole-brain CEST imaging using 3D-EPI at 7T

Suzan Akbey¹, Philipp Ehses¹, Rüdiger Stirnberg¹, Moritz Zaiss², and Tony Stöcker^1,3

¹German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany, ²Max-Planck Institute for Biological Cybernetics, Tübingen, Germany, ³Department of Physics and Astronomy, University of Bonn, Bonn, Germany

Synopsis

Compared to conventional single-slice or small volume measurements, whole-brain CEST acquisitions present new challenges, in particular at ultra high-field (UHF). In a first step to account for the broader B1 variation, we analyzed the distribution of reference voltages in the brain of nine subjects. Based on these results, we repeated a whole-brain CEST experiment with seven different reference voltages to optimize the correction for B1 inhomogeneities in the CEST contrast maps. Additionally, we extended the saturation-offset list to compensate for the higher variation of the static magnetic field for whole-brain experiments compared to examinations with smaller FOVs.

Introduction

Chemical exchange saturation transfer (CEST) imaging provides information about the molecular microenvironment¹ which may offer new insights into neurodegeneration. We aim to observe CEST throughout the whole brain since neurodegenerative diseases are not restricted to a certain brain area.

Whole-brain acquisitions have to deal with a broader B0 distribution and stronger B1 variations compared to smaller FOVs. To identify challenging brain regions for CEST quantification and as a first step in the development of an improved B1 correction, we analyzed the B1 distribution in nine subjects. CEST with a 3D-EPI readout was employed² in order to acquire more saturation pulse power (B1) levels in a reasonable scan time. Number and range of saturation levels were optimized for whole-brain CEST. Additionally, we extended the frequency-offset list to compensate for larger B0 inhomogeneities.

Methods

Human in-vivo data were acquired on a 7T MRI (Siemens) using a 32/1 (Rx/Tx) channel head coil (Nova medical). B1-Mapping was performed using the 3DREAM³ sequence on nine subjects. The resulting flip angle maps were converted to reference voltages (U_ref) (defined by the voltage required for full inversion with a 1ms block pulse) and then registered to the MNI space⁴. We compared U_ref across the whole brain to a single slice (1mm) and to a small volume (50 slices) and analyzed U_ref for different cortical regions.

The readout module consists of centric-reordered 3D-EPI readout with water excitation² (resolution=2.0x2.0x2.0mm³, sagittal FOV=212x160x212mm³, TR_par=18.35ms, TE =5.9ms, t_readout=734ms, TR=4.33s, TA=4:36min, BW=2246Hz/Px, FA_nominal=10˚, GRAPPA 3x2, partial Fourier=6/8x1).

CEST saturation⁵ consists of a train of 120 Gaussian-shaped RF pulses applied before each volume acquisition (t_pulse=15ms, t_sat=3.6s, duty-cycle=50). The frequency-offset list was expanded from 52 to 58 entries between -6 and 6 ppm in order to increase robustness against B0 inhomogeneity. We used a nominal B1_sat of 0.8uT and repeated the measurement seven times with U_ref in the range of 150-450V. Additionally, we use the assumption of zero-crossing of all CEST effects at 0V.

Z-spectra and CEST-maps were obtained from a Lorentzian fit⁵ after motion/distortion correction⁴ and B0/B1-correction^5-7 using B0 (GRE-based field mapping) and B1 (3DREAM³) maps, respectively. The homogeneity of MTR_Rex maps was analyzed in white and gray matter regions identified from a T1-weighted scan.

Results

Figure 1a shows the local distribution of U_ref across the brain. Compared to single-slice or small volume experiments, the histogram of whole-brain examinations is broader (Fig. 1b). Lower regions of the brain, e.g. the cerebellum, require a higher U_ref than regions in the cerebrum (Fig. 2).

Uncorrected and 1-point-corrected MTR_Rex maps for MT (Fig. 3) show a strong correlation to B1 whereas the maps corrected with two or more reference voltages are more homogeneous and the distribution of MTR_Rex for MT, NOE and APT is narrower (Fig. 4). MTR_Rex of the 2-point-corrected data is slightly shifted to smaller values.

Extending the number of offsets improves the quality of the CEST maps in the regions of high B0-deviation (Fig. 5). In regions of very low B1, CEST quantification fails.

Discussion and Conclusion

Two anatomical regions, cerebellum and temporal lobe, are particularly challenging for quantification due to the requirement of high U_ref. Additionally, the required U_ref in these regions vary stronger between subjects. B1 shimming or parallel transmit excitation may improve CEST quantification in the cerebellum and temporal lobe.

Accounting for the range of B1 values, we chose U_refup to 450V without exceeding SAR limits. This represents a reasonable upper limit for CEST quantification with 2-7-point B1-correction (compare Fig. 4). Our results indicate that acquiring three reference voltages in the range 200-450V is a feasible option for whole-brain CEST. It is slightly more stable against stronger B1 variations than the 2-point B1-correction in which we observed systematic CEST contrast differences with a yet unknown origin, which are subject of further investigation. In future work, parallel transmit may help to reduce the number of B1 sampling points required for accurate quantification⁸.

The distribution of B0 across the brain is larger than ±1ppm, prolonging the time-requirement of a WASSR⁹ acquisition. A GRE-based fieldmap is a more time-efficient alternative that was chosen here. However, depending on the choice of echo-times, this may require phase-unwrapping. Nevertheless, the quantification of CEST effects in regions of high ΔB0 is challenging and may be exacerbated by a low B1 field in the affected regions.

In this work, we showed that 3D-EPI is a feasible option for whole-brain CEST at UHF. However, in order to allow accurate quantification, strong B0 and B1 inhomogeneities associated with this whole-brain approach need to be accounted for.

Acknowledgements

SA holds a PhD Fellowship from the Konrad Adenauer Stiftung.

The financial support for MZ of the Max Planck Society, German Research Foundation (DFG, grant ZA 814/2-1), and European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 667510) is gratefully acknowledged.

References

1. van Zijl PCM, Lam WW, Xu J, et al. Magnetization Transfer Contrast and Chemical Exchange Saturation Transfer MRI. Features and analysis of the field-dependent saturation spectrum. NeuroImage. 2017; 04:1053-8119

2. Akbey S, Ehses P, Stirnberg R, et al. Single-shot whole-brain CEST imaging using centric-reordered 3D-EPI. Proc. Intl. Soc. Mag. Reson. Med. 2018; 26:2231

3. Brenner D, Tse DHY, Pracht ED, et al. 3DREAM – A Three-Dimensional Variant of the DREAM Sequence. Proc. Intl. Soc. Mag. Reson. Med. 2014; 22:1455

4. Smith SM, Jenkinson M, Woolrich MW, et al. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage, 2004; 23:208-219.

5. Zaiss M, Windschuh J, Paech D, et al. Relaxation-compensated CEST-MRI of the human brain at 7T: Unbiased insight into NOE and amide signal changes in human glioblastoma. NeuroImage. 2015; 112:180-188.

6. Windschuh J, Zaiss M, Meissner JE, et al. Correction of B1-inhomogeneities for relaxation-compensated CEST imaging at 7 T. NMR in Biomed. 2015; 28(5):529-537.

7. Zaiss M. CEST sources. www.cest-sources.org. Accessed November 5, 2017.

8. TSE DMY, da Silva NA, Poser BA, Shah NJ. B1+ Inhomogeneity Mitigation in CEST Using Parallel Transmission. Mag. Reson. Med. 2017; 78:2216-2225

9. Kim M, Gillen J, Landman BA, et al. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Mag. Reson. Med. 2009; 61(6):1441-1450.

Figures

Figure 1: a) Axial, coronal and sagittal slices of the mean reference voltage map (N=9) in MNI-domain. The axial slice (1mm) is marked in purple in the sagittal view. The related normalized single slice histogram is shown in b) compared to the histogram of a small volume (50 slices, in red) and the distribution of the reference voltages across the whole brain (blue). The larger the volume is, the longer the tail of the reference voltage distribution. This necessitates high power measurements, which are restricted due to the SAR limits.

Figure 2: a) Maximum probability map (FSL³) to differentiate the brain regions in MNI-space. b) The table shows the mean reference voltage within a brain region over all subjects as well as the standard deviation within the regions averaged over the subjects and the standard deviation between subjects averaged over the region. Cerebellum and temporal lobe show the highest mean reference voltages as well as the largest variation between subjects and within the regions.

Figure 3: CEST maps of the contrast parameter MTR_Rex(δ_MT) of representative axial, coronal and sagittal slices of a healthy volunteer. Uncorrected (B1_nom=0.8uT) and Z-B1-corrected (B1=0.9uT) image data calculated with increasing number of B1 sampling points (1point [300V], 2point [200V,400V], 3point [200V,300V,450V], 4point [150V,200V,300V,450V], 7point [150V,200V,250V,300V,350V,400V,450V]). A decreasing impact of B1 inhomogeneity is observed for all B1-corrected maps, with increased image quality for higher numbers of B1-values. 2-point corrected data differs mostly in the lower regions of the brain and shows a global systematic offset compared to the other datasets. Strong ΔB0 hampers the CEST quantification in all images.

Figure 4: Histograms of MTR_Rex (δ_MT=-1ppm, δ_NOE=-3.5ppm, δ_amide=3.5ppm) in ROIs representing white and gray matter of a healthy subject for different numbers of B1 sampling points included in the correction. The distributions of all contrasts are much narrower for 2-point to 7-point B1-corrected data. The 2-point B1-corrected data is slightly shifted to smaller MTR_Rex values (except for the NOE in GM).

Figure 5: Sagittal slice of the MTR_Rex contrast (δ_MT=-1ppm, δ_NOE=-3.5ppm, δ_amide=3.5ppm) for 52 and 58 offsets between -6 and 6ppm as well as the related B0 and B1maps. The image quality of the CEST maps improves in the regions of higher B0 deviations (red arrows) for higher number of saturation offsets. In both cases, the quantification is challenging in very low B1 regions (green arrow).

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)

4019