A correction to the bandwidth difference method for assessment of magnetic resonance static field homogeneity

Abstract

Abstract Background Published guidance offers four methods of static field inhomogeneity (ΔB0) measurement, including a bandwidth difference method. However, an error exists in the implementation of this method, which we demonstrate analytically and experimentally. Purpose To propose and validate a correction to the bandwidth difference method for quantifying ΔB0. Methods We outline a correction in the method for quantifying local ΔB0 by measuring spatial translations rather than diameter measurements between MRI images with different receive bandwidths. To validate the approach, we acquired images of a cylindrical phantom containing a fixed array of fluid‐filled pins. We adjusted static field shims to create approximately linear and quadratic field perturbations and acquired high‐ and low‐bandwidth spoiled gradient echo sequences with anterior‐posterior (AP) and left‐right (LR) frequency‐encoding. We then measured translations of centroid pin locations between low‐ and high‐bandwidth images. Finally, we propose and validate an implementation strategy using a spherical phantom. In both cases, a vendor‐provided field mapping sequence provided reference homogeneity measurements. Results Under manually adjusted shim settings, root‐mean‐square (RMS) and peak‐to‐peak (PP) ΔB0 were 2.46 and 6.29 parts per million (PPM) respectively, measured with the field mapping technique. Results of the proposed bandwidth difference method showed strong agreement; RMS and PP ΔB0 were 2.38 and 6.22 PPM respectively for AP frequency‐encoding and 2.38 and 6.08 PPM respectively for LR frequency‐encoding. In a spherical phantom, the proposed method agreed with reference measurements to within 0.25 PPM (< 7%). However, PP ΔB0 was underestimated using the diameter‐based method—in a spherical phantom, measured ΔB0 was 0.74 PPM compared to 3.52 and 3.76 PPM using the reference and revised methods, respectively. Conclusions This work experimentally validates a revised bandwidth difference approach for ΔB0 quantification, showing that it remains sensitive under spatially nonlinear ΔB0; the diameter‐based method is shown to be inaccurate for such cases.