Recovery from Errors Due to Domain Truncation in Magnetic Particle Imaging: Approximation Error Modeling Approach

In magnetic particle imaging, the concentration of magnetic nanoparticles is imaged based on their nonlinear magnetization response to an applied magnetic field. Because of security limitations regarding the magnetic field, the field of view does not always cover the entire body being imaged. However, nanoparticles outside the field of view can also be excited by the applied magnetic field and therefore contribute to the measured signal. This leads to domain truncation artifacts, which can often be observed in the reconstructed particle concentrations. In this paper, a computational method for reducing such truncation artifacts is proposed. For this aim, an approximative statistical model for the modeling errors caused by domain truncation is constructed, and the reconstruction problem in magnetic particle imaging is solved in the Bayesian framework of inverse problems. Several numerical test examples illustrate the successful recovery from truncation artifacts.