Recent Progress in Model-Based Reconstruction using Approximate Particle Models and low Calibration Effort

Abstract

In magnetic particle imaging (MPI) one of the key challenges is still the model-based reconstruction problem particularly when using multi-dimensional Lissajous-type excitation. Already during the last IWMPI 2023 promising results towards a solution to this key problem have been presented from different groups. A combination of realistic physical modeling of the nanoparticles' magnetization behavior and a careful calibration of the remaining scanner components result in improved model-based reconstructions when compared to the fully calibrated system matrix approach. In addition, the question has been asked to which extent the equilibrium model can be extended to include particle anisotropy, which significantly increases the accuracy of the model while still keeping the computational effort low. The previous findings from different groups at the last IWMPI provided the starting point for a larger joint collaboration where we extended the previous results by an extensive series of experiments which showcase that the model-based reconstruction can be applied in aligned and unaligned immobilized cases but even more important also to fluid cases. In this IWMPI contribution we present an update on the ongoing works in the consortium addressing the key problem of model-based reconstruction in MPI for multi-dimensional Lissajous-type excitation. We showcase the accuracy and the computational effort for model evaluation on different system matrices and various particle phantoms being measured recently on the Bruker preclinical MPI system.