We develop an efficient edge preserving and noise reducing reconstruction method for MPI using a nonnegative fused lasso model. We devise a discretization adapted to the acquisition geometry of a preclinical MPI scanner and an efficient solver for that setup. Our prototype implementation processes a 3D volume within a few seconds. We demonstrate the improvement in reconstruction quality over the state-of-the-art method in an experimental medical setup for in-vitro angioplasty.
This presentation is part of Minisymposium “MS56 - Mathematical and Computational Aspects in Magnetic Particle Imaging (2 parts)”
organized by: Tobias Kluth (University of Bremen) , Christina Brandt (University of Hamburg) .