A Distributed Compressive Sensing Strategy for Non-Cartesian MRI: Applications to SWIRLS CE-MRA

Joshua D. Trzasko¹, Yunhong Shu², Armando Manduca¹, John Huston III², Matt A. Bernstein²

In non-Cartesian image reconstruction, equality data constraints cannot be explicitly enforced, which limits the applicability of popular methods like projection-onto-convex-sets (POCS) to this area. In this work, we demonstrate that they can be implicitly enforced via a specific affine projection, and discuss numerical methods for efficiently imposing them. We use this construction to develop an efficient Compressive Sensing reconstruction based on block-wise redundant sparsity constraints, which results in strong reconstruction performance. We demonstrate the proposed reconstruction strategy for CE-MRA acquired using the 3D SWIRLS trajectory.