Randomized Greedy Kaczmarz for Multi-Contrast Magnetic Particle Imaging

Multi-contrast magnetic particle imaging allows for separating signals from different tracer materials or microenvironments from one joint measurement, yielding multi-channel images that distinguish tracers or encode properties such as temperature or viscosity. However, leakage artifacts enforce a noticeably longer application of iterative standard methods like the Kaczmarz algorithm to reach satisfactory reconstruction results. A greedy randomized version of the Kaczmarz method is adapted, showing the potential for faster recovery.