Fan Lam1,2, Raman Subramanian3,
Dan Xu3, Kevin F. King3
1Electrical & Computer
Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United
States; 2Beckman Institute, University of Illinois at
Urbana-Champaign, Urbana, IL, United States; 3GE Healthcare,
Waukesha, WI, United States
We present a novel reconstruction scheme incorporating not only the prior information that the MR image is sparse in certain transformation domain but also the support information for the target image to be reconstructed. Support can be detected either from low resolution estimate or from certain transformation domain of a high resolution reference image. A mix weighted L1-L2 regularization formulation is established for reconstruction. Data from a noncontrast MRA and a brain imaging experiment are used to demonstrate the advantageous performance of the proposed method compared to conventional compressed sensing based reconstruction from sparsely sampled data.