Super resolution estimation of B0 field map of brain MRI using model-based methods employing intrinsic priors

B0 inhomogeneity can cause image distortion and signal loss, especially for MR images acquired at long echo times. Image artifacts induced by B0 inhomogeneity could potentially be corrected by B0 mapping. In spite of previous works on B0 mapping, acquiring high-resolution B0 field map in a short time remains challenging. On the other hand, only accurate high-resolution B0 field map would be useful in correction of intravoxel field inhomogeneities. In this work, we proposed a super resolution B0 estimation method to improve the B0 mapping near air/tissue interface, which is based on the smooth prior of B0 field and the Fourier relationship between high-resolution and low-resolution B0; we also compared it with B0 estimation methods that employ structural prior of the brain by susceptibility voxel convolution, and another method leveraging brain B0 template derived from a cohort of 222 subjects. We found that the proposed method with intrinsic prior performed better than susceptibility voxel convolution and template based methods.