Exploiting parameter sparsity in model-based reconstruction to accelerate proton density and T2 mapping

Abstract

T2 mapping is a powerful noninvasive technique providing quantitative biological information of the inherent tissue properties. However, its clinical usage is limited due to the relative long scanning time. This paper proposed a novel model-based method to address this problem. Typically, we directly estimated the relaxation values from undersampled k-space data by exploiting the sparse property of proton density and T2 map in a penalized maximum likelihood formulation. An alternating minimization approach was presented to estimate the relaxation maps separately. Both numerical phantom and in vivo experiment dataset were used to demonstrate the performance of the proposed method. We showed that the proposed method outperformed the state-of-the-art techniques in terms of detail preservation and artifact suppression with various reduction factors and in both moderate and heavy noise circumstances. The superior reconstruction performance validated its promising potential in fast T2 mapping applications.