High resolution gradient-echo EPI using a shim insert coil at 7T: Implication for BOLD fMRI.
Tae Kim1, Yoojin Lee1, Tiejun Zhao2, Hoby Hetherington1, and Jullie Pan1

1University of Pittsburgh, Pittsburgh, PA, United States, 2Siemens Medical Solution USA, INC, Pittsburgh, PA, United States

Synopsis

High degree/order shimming was applied to improve field homogeneity thereby reducing susceptibility-induced distortion in high resolution gradient-echo EPI using a shim insert coil at 7T. Use of the shim insert improved the overall homogeneity across the entire brain by 30% in comparison to conventional 1st&2nd degree/order shimming. The susceptibility-induced displacement improved by more than 2cm and the number of activated pixels increased by 150% with higher degree shimming in regions such as the anterior temporal and frontal lobes. Our study demonstrates that the use of higher order/degree shims improves GE-EPI BOLD signal at high field.

purpose

B0 inhomogeneity from differences in magnetic susceptibility is well known to result in signal reduction and spatial distortion in gradient echo BOLD echo planar imaging (GE-EPI) which increases linearly with field strength. However, the extent to which improved B0 shimming can reduce these effects has not been quantitatively evaluated. In this report we evaluate the extent by which 1st–4th degree/order shimming improves GE EPI BOLD imaging at 7T.

Method

8 healthy volunteers were studied at 7T (Siemens Magnetom PTX) using an 8x2 (8coils/row x2 rows) transceiver array. The array was divided into 8 pairs of coils with each circumferential pair driven by a single independent RF channel using a 45° splitter. RF shimming was performed for optimum global homogeneity. A 38cm ID shim insert coil consisting of 3rd, 4th degree and two 5th degree shims with 10A shim supplies (Resonance Research Inc.) was used for high degree shimming. B0 mapping was performed using a 5 time point (1.0 to 8ms evolution times) multi-slice measurement with shim values calculated using a non-iterative least squares algorithm (Bolero, Bo loop encoded readout) [1]. Single-shot GE-EPI was acquired with FOV=19.2x19.2 cm2, matrix=96x96, 2mm isotropic voxels, TR/TE=3.5 s/24 ms, FA=60° and 48 slices. To induce a whole brain BOLD response, a 21sec breath-hold protocol was performed five times using a block-design paradigm (140s−[(21s)−35s]x5; parentheses indicates breath-hold period). Each run was repeated twice for 1st&2nd and 1st-4th degree/order shimming. The B0 induced voxel displacements were calculated by SPM. All data were processed with AFNI and FSL. High resolution (0.75mm3) MP2RAGE images were used for anatomical registration [2]. tSNR maps were obtained from the pre-breath-hold time series. BOLD maps were estimated by fitting the least squares of the linear regression using AFNI (3dDeconvolve) following normalization by the mean signal intensity at each voxel. Activation was determined on a pixel-by-pixel basis (p-value < 0.05). To assess the incremental effect of improved B0 shimming achieved with 1st-2nd degree shims versus 1st-4th degree shims i.e., [B01&2degree| - |B01-4degree|, eleven bins of improving B0 homogeneity were generated in 10Hz ΔB0 increments, i.e., 0–10Hz (representing a small amount of B0 improvement), 11–20 Hz, … , 91–100 Hz, >100Hz (a very large amount of B0 improvement). T1W images and bins were warped into the EPI domain using the field maps shimmed with either 1st&2nd degree vs 1st-4th degree shimming.

Results

From 8 subjects, the standard deviation of B0 over the entire brain with higher order shimming decreased 26% (29.9 ± 4.5 Hz to 22.1 ± 3.8 Hz) for 1st&2nd vs. 1st–4th degree shimming. Fig. 1A shows B0 maps obtained from 1st&2nd and 1st–4th degree shimming. Fig. 1C demonstrates distorted T1W images for each shimming condition by forward warp from undistorted T1W images (Fig 1B). With 1st & 2nd degree shimming, the very high susceptibility areas (seen particularly in the inferior regions) of the T1W image disappeared due to huge voxel displacement. With 1st-4th degree shimming, this loss is much less. Fig. 1D shows corresponding EPI images, which agrees well with the forward warped images in Fig. 1C. tSNR and BOLD maps (Fig. 2A,B) for 1st&2nd vs. 1st–4th degree shimming are shown. With 1st-4th degree shim, the tSNR and the number of activated pixels in the high susceptibility areas are increased in comparison with 1st&2nd shimming. To visualize the regional improvement in B0 homogeneity, pixels with 10Hz ΔB0 increments are overlaid on the anatomical image (Fig. 3A). The voxel displacement due to the field inhomogeneity decreased ~3 times with 1st–4th degree shim as compared with 1st & 2nd shim (Fig. 3B, voxel displacement improvement = 0.08xΔB0 (Hz) improvement, R2=0.93). The number of activated pixels and tSNR summation was largely improved with 1st–4th degree shim due to less distortion (Fig. 3C).

Discussion

Although high resolution acquisition is helpful to reduce the signal dropout, in high susceptibility regions the increased voxel displacement induces pixel movement beyond the field of view or into overlapped pixels. This effect cannot be corrected with post-processing and consequently decreases the number of activated pixels. While high degree shimming improved the mean tSNR value by a relatively small ~10%, its major effect is to minimize pixel distortion, thus retaining a higher fraction of all pixels and improved activation.

Conclusion

High degree shimming allows substantial enhancement of BOLD activation at 7T, especially in high susceptibility regions of the brain.

Acknowledgements

NIH grant, R01EB011639; R01NS081772; R01NS090417; R01EB009871 and RSNA grant, RSCH1314

References

1. Hetherington HP, Chu WJ, Gonen O, Pan JW. Robust fully automated shimming of the human brain for high-field 1H spectroscopic imaging. Magn Reson Med 2006;56(1):26-33.

2. Marques JP, Kober T, Krueger G, van der Zwaag W, Van de Moortele PF, Gruetter R. MP2RAGE, a self bias-field corrected sequence for improved segmentation and T1-mapping at high field. Neuroimage 2010;49(2):1271-1281.

Figures

Distortion effect on 1st&2nd and 1st–4th degree shimming. (A) B0 maps. (B) undistorted T1-weighted images. (C) distorted T1-weighted images generated by forward warp using the corresponding field maps. (D) corresponding 2mm isotropic EPIs. Larger distortion was observed with 1st&2nd shim as compared with 1st-4th degree shim.

Four slices are demonstrated for the comparison of tSNR (A), and BOLD activation (B) maps from one subject. The lesser tSNR and BOLD activation are evident with the 1st-2nd degree shimmed condition.

(A) ΔB0-bins are overlaid on the anatomical image. (B) The voxel displacement was linearly increased with ΔB0-bin within this range. The displacement with 1st–4th shim was much less than those with 1st&2nd. (C) The number of activated pixels was improved after 40Hz of B0 improvement. Colorbars: each range of ΔB0bin



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3719