ACCURATE T2 MAPPING WITH SPARSITY AND LINEAR PREDICTABILITY FILTERING

T2 mapping provides a quantitative manner to access tissue structure, composition, water content and iron levels. Nevertheless, due to the relative long scanning time, its practical usage is limited. This paper addresses this problem using a novel iterative nonlinear filtering method to achieve sparse sampling reconstruction. Specifically two filters are involved. One is the soft thresholding operator promoting spatial sparsity and temporal redundancy. The other is the Hankel matrix low-rank approximation enforcing the exponential structure along echo time dimension. The proposed method has been validated based on a brain T2 experiment data, and is shown to provide high image quality.