Purpose

The improvement of B₀ shimming is important for gradient echo functional MRI where T₂^* signal loss and precession occur from both physiological and non-physiological susceptibility effects, these effects being substantially worse at 7T and with echo-planar imaging where pixel shift produces complex signal distortion. As a result, there has been much effort on methods to improve GE fMRI. Here we describe the quantitative improvement in whole brain BOLD signal and activated pixel counts using high degree spherical harmonic shimming.

Methods

B₀ shim insert, field map and shimming: A shim insert coil consisting of 3^rd, 4^th degree and two 5^th degree shims (ZC4, ZS4) with 10A shim supplies (RRI, Billerica, MA) was used for high degree shimming, denoted as 1^st-4^th+ shimming. The B₀ field was mapped using the B₀ loop encoded readout (Bolero) method, with shimming performed by a least squares optimization based on a calibrated spherical harmonic model with a user-determined target ROI shown in Fig. 1A. To analyze EPI data, all pixels were binned in 10Hz increments according to change in |B₀| offset between 1^st – 2^nd and 1^st – 4^th+ shims, i.e., ΔB₀=|B₀^1,2| - |B₀^1-4+| (improved homogeneity pixels with 1^st-4^th+ compared to 1^st-2^nd shim are in positive bins; deteriorated pixels in negative bins). RF coil: An 8x2 (two rows, eight coils/row) inductively decoupled elliptical transceiver array was used in parallel transmit mode. RF shimming per subject was performed with eight 1kW transmit channels which were split and phased to generate a circularly polarized mode-like distribution. The achieved B₁⁺ was 460±51Hz over the entire head at 1140±113 Watts total applied power. EPI and breath hold BOLD activation: Single-shot GE-EPI 2mm³ (48 slices, matrix size=96´96, bandwidth/pixel=13.88Hz, echospacing 750ms) was acquired in 6 volunteers; FOV=19.2x19.2cm2, TR=3.5s, TE=24ms, phase partial Fourier=6/8, FA=60°. For the whole brain BOLD signal response, a 21-sec breath-hold protocol was performed five times using a block-design paradigm (140s − [21s (breath hold) −35s] x5). N=8 subjects were asked to exhale prior to breath-holding to preclude a biphasic change (1). Each run was acquired with 1^st-2^nd and 1^st-4^th+ degree shimming and signal averaged in randomized order. Processing was performed using AFNI (http://afni.nimh.nih.gov/afni/), FSL (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) and SPM (http://www.fil.ion.ucl.ac.uk/spm/) programs combined with in-house programs written in Matlab R2013b. The Bolero B0 map was used to calculate inhomogeneity induced voxel displacements. Following normalization by the mean signal intensity at each voxel, BOLD maps were estimated by linear regression using AFNI (3dDeconvolve), which also included motion parameters. The significance of activation was determined on a pixel-by-pixel basis over each DB0 bin(p-value<0.05). MP2RAGE images were used for anatomical registration using (0.75 mm)³ resolution with TR=6s, TE=3.1ms, TI=0.8/2.7s, and GRAPPA 3. All studies were approved per the University of Pittsburgh IRB.

Results

The ΔB₀ bins, as presented in Talairach space from n=8 volunteers (Fig. 1B) shows that the high degree shimming had its greatest effects over the inferior frontal and temporal regions. For both the positive and negative bins, the BOLD % signal change is significant, being increased in the positive bins and decreased in the negative bins. In both positive and negative bins, there are significant changes in numbers of activated pixels. From n=8 volunteers, Fig. 2A shows that the high degree shimming significantly increased the numbers of activated pixels in the positive Δ|B₀| bins, particularly above +40Hz, while they decreased in the negative bins. For all bins, the activated pixel numbers (green) were significantly different between 1^st-2^nd and 1^st-4^th+ shims. For most of the bins, the BOLD %signal change (orange) was significantly different (p<0.05) between the 1^st-2^nd and 1^st-4^th+ shims; Fig. 2B shows the locations and fractional change in activated pixel numbers and BOLD % signal change averaged over the 8 volunteers shown in Talairach space. Data from a single subject is shown in Fig. 2C.

Discussion

As an independent means of improving field homogeneity, high degree shimming does not interact with other imaging parameters and is compatible with whole brain coverage necessary for simultaneous multi-slice methods. Comparing 1^st-2^nd and 1^st-4^th+ shimming, the Δ|B₀| changes spatially vary over the brain with the majority of the brain either maintaining or increasing signal intensity and activation. Focusing on the pixels that substantially change (either improved or worsened, outside of the two center Δ|B₀| bins) with the 1^st-4^th+ shimming, for each 1 activated pixel that is lost, there are 7.9 activated pixels gained. The largest regions of activation increase are seen in the inferior frontal region while the largest regions of activation decrease are in the middle temporal lobe. Overall, there is a 4.3% increase in total activated pixels.

Acknowledgements

This work is supported by NIH R01EB011639; R01NS081772; R01NS090417; R01EB009871 and RSNA RSCH1314.

References

1. Li TQ, Kastrup A, Takahashi AM, Moseley ME. Functional MRI of human brain during breath holding by BOLD and FAIR techniques. Neuroimage 1999;9(2):243-249.