Multi-Contrast CSMRI Using Common Edge Structures with LiGME Model

CSMRI is the high-speed magnetic resonance imaging (MRI) technique using the compressed sensing (CS) theory. Based on the fact that multiple MR images of different contrasts, e.g., T1-weighted and T2-weighted images, are scanned in clinical practice, Ehrhardt et al. proposed multi-contrast CSMRI utilizing the edge information of a different contrast image obtained from the full-sampling k-space data. In this paper, we propose to extend the method of Ehrhardt et al. to the linearly involved generalized Moreau enhanced (LiGME) model. Since a directional total variation based on the edge information becomes closer to a group
$\ell_{0}$
pseudo-norm by introducing the LiGME model, we will be able to reconstruct large edges more accurately. Simulations using actual MR images demonstrate the effectiveness of the proposed method.