レベル集合制約を用いた圧縮センシングMRI

抄録

CS-MRI is a high-speed magnetic resonance imaging technique based on compressed sensing theory. It is known that the utilization of dictionary learning (DL) for CS-MRI leads to better image reconstruction results. In the optimization problem adopted by the DL-based CS-MRI, we must appropriately adjust the weights of data fidelity term and dictionary fidelity term, respectively, for each MR image. In this paper, we formulate a novel optimization problem where the above two fidelities are expressed as level set constraints, and solve this optimization problem by using alternating direction method of multipliers (ADMM). Numerical experiments show that the proposed method can reconstruct several MR images with high accuracy by using the same parameters.