On the Impact of Artifacts Induced by Mismatches Between Auto-Calibration Signal and Accelerated 3D GRE Data at 11.7T.

PURPOSE: The study aims at investigating B 0 $$ {B}_0 $$ field inhomogeneity artifacts arising from remote locations in the FOV and encountered in accelerated 3D gradient-recalled echo (GRE) sequences at ultra-high field, and at providing mitigation strategies. METHODS: Measurements were conducted at 11.7T using a head-shaped phantom and an accelerated 3D GRE sequence with either integrated or external auto-calibration signal (ACS) lines. Simulations were performed to reproduce the artifacts. The effects of varying GRAPPA reconstruction parameters (kernel size and regularization) were also examined. RESULTS: B 0 $$ {B}_0 $$ field inhomogeneities located outside the B 0 $$ {B}_0 $$ shimmed region of interest (i.e., the brain) were observed to return ripple-like artifacts within this region, particularly at long echo times. The simulation results support these findings, and the idea that the observed artifact originates from a mismatch between ACS and accelerated data due to intra-voxel dephasing at different resolutions (ACS lines having an intrinsically lower resolution). The short echo time enabled by external (i.e., preacquired) ACS lines reduced artifacts compared to integrated ones. Varying GRAPPA kernel sizes and increasing the number of ACS lines can improve image quality, yet without full compensation. CONCLUSION: This study highlights ripple-like artifacts amplified with field strength and arising from a lack of coherence between the ACS and imaging (3D GRE) signal caused by B 0 $$ {B}_0 $$ intra-voxel dephasing. To minimize these artifacts, care should be taken in order to preserve the relevant information in the ACS data to properly compute the GRAPPA kernels.