A method for macroscopic B0 field inhomogeneity compensated SWI using 3D z-shim multi-echo GRE

Purpose Susceptibility weighted imaging (SWI) based on GRE provides enhancement of susceptibility difference which is useful for visualization of vein, micro-hemorrhage and iron deposition. However, GRE suffers from the macroscopic B0 inhomogeneity due to air/tissue boundary and this effect is shown as SNR loss both in magnitude and phase. Previously, 3D z-shim multi-echo GRE (mGRE) pulse sequence demonstrated that z-shimming can successfully recover both magnitude and phase SNR loss (Fig. 1 red arrow) due to B0 field inhomogeneity. Here, an improved algorithm for B0 field inhomogeneity compensated SWI, applicable to both magnitude and phase, is proposed using the 3D z-shim mGRE sequence. Methods The 3D z-shim mGRE sequence collects both conventional echoes (odd echoes) and z-shimmed echoes (even echoes) in a single-scan. Using these echoes, both magnitude and phase data without field inhomogeneity were generated by the proposed algorithm shown in Fig. 2. Phase processing To acquire B0 field inhomogeneity compensated field map, a phase combine method between conventional and z-shimmed echo was used. Background field removal method (PDF) was applied to acquire local field map (LFM). Then, local phase value at specific TE (echo time) was estimated using equation described in Fig. 2 (γ: gyromagnetic ratio, B0 : 3 [T]). This local phase avoids wrapping artifact in SWI process (Fig. 3 phase mask). Magnitude processing The field map obtained by the previous process is not only useful for acquiring accurate local field information but also for R2 fitting. This is because magnitude loss due to B0 field inhomogeneity can be modeled as a function of the field inhomogeneity (e.g. voxel spread function (VSF) method). The VSF method which is appropriate for 3D mGRE was modified to incorporate the z-shimmed echo signals and was applied for R2 fitting to acquire M0 and R2 values. Afterwards, macroscopic B0 field inhomogeneity compensated magnitude was estimated using a monoexponential model as shown in Fig. 2.