A sparse row-action algorithm for magnetic particle imaging

The image reconstruction in Magnetic Particle Imaging (MPI) relies on efficiently solving an ill-posed inverse problem. Current state-of-the-art reconstruction methods are either based on row-action methods with fast convergence but limited noise suppression or advanced sparsity constraints showing better image quality, but suffering from a higher computational complexity and slower convergence. In this contribution, we propose a novel row-action framework where advanced sparsity constraints, e.g., a combination of l₁-and TV-norm, can be included. Its performance is numerically evaluated on simulated and real MPI data, showing a significant reduction of computation time while retaining the enhanced imaging quality.