Patryk Szwargulski, M.Sc.
Universitätsklinikum Hamburg-Eppendorf (UKE)
Sektion für Biomedizinische Bildgebung
2ter Stock, Raum 203
Technische Universität Hamburg (TUHH)
Institut für Biomedizinische Bildgebung
Gebäude E, Raum 4.044
Am Schwarzenberg-Campus 3
- Magnetic Particle Imaging
- Image Reconstruction
- Signal and Image Processing
In 2015 Patryk Szwargulski graduated with a master's degree thesis on Fast Reconstruction of Magnetic Particle Imaging Data using the Focusfields. Currently he is a PhD student in the group of Tobias Knopp for experimental Biomedical Imaging at the University Medical Center Hamburg-Eppendorf and the Hamburg University of Technology.
|Title: Simultaneous imaging of widely differing particle concentrations in MPI: problem statement and algorithmic proposal for improvement|
|Written by: M. Boberg, N. Gdaniec, P. Szwargulski, F. Werner, M. Möddel, and T. Knopp|
|in: Physics in Medicine & Biology 2021|
Note: article, artifact
Abstract: Magnetic Particle Imaging (MPI) is a tomographic imaging technique for determining the spatial distribution of superparamagnetic nanoparticles. Current MPI systems are capable of imaging iron masses over a wide dynamic range of more than four orders of magnitude. In theory, this range could be further increased using adaptive amplifiers, which prevent signal clipping. While this applies to a single sample, the dynamic range is severely limited if several samples with different concentrations or strongly inhomogeneous particle distributions are considered. One scenario that occurs quite frequently in pre-clinical applications is that a highly concentrated tracer bolus in the vascular system "shadows" nearby organs with lower effective tracer concentrations. The root cause of the problem is the ill-posedness of the MPI imaging operator, which requires regularization for stable reconstruction. In this work, we introduce a simple two-step algorithm that increases the dynamic range by a factor of four. Furthermore, the algorithm enables spatially adaptive regularization, i.e. highly concentrated signals can be reconstructed with maximum spatial resolution, while low concentrated signals are strongly regularized to prevent noise amplification.