Magnetic Field Characterization

Each Magnetic Particle Imaging Scanner has individual magnetic field profiles with distortions that cannot always be simulated.  In particular, eddy current distortions caused by conductive material in the vicinity of the field generating coils is difficult to simulate. Many MPI areas benefit from characterization of the magnetic field: The most basic advantage is to get a rough estimate for the field-of-view (FOV), as a combination of the drive field and the gradient field. From a hardware point of view, the drive-field profile can be used to manufacture optimized dedicated receive coils, as in "Design of a head coil for high resolution mouse brain perfusion imaging using magnetic particle imaging". For a successful model-based reconstruction that utilizes an artificial generated system matrix, the input parameters to generate the system matrix are particularly important. In "Model-based Calibration and Image Reconstruction with Immobilized Nanoparticles" it was shown that besides a robust particle model and the MPI transfer function, accurate magnetic field values are necessary to describe the scanner dependent measurement signal. Additionally, knowledge about the magnetic field profiles can be used to reduce the number of multi-patch system-matrix measurements as in "Generalized MPI Multi-Patch Reconstruction using Clusters of similar System Matrices".

Generally, the magnetic fields in MPI can be separated into "static" and "dynamic" fields. Other than the "static" fields (gradient and focus fields) that can be measured by a Hall probe, the "dynamic" fields (drive fields) can be measured using a coil sensor and Faradays law of induction. To obtain accurate field profiles, the FOV can be measured in form of a system matrix. However, since magnetic fields satisfy the Laplace equations, they can be expressed within a sphere as a series of spherical harmonic functions by integrating solely over the surface of the sphere. Efficient determination of the coefficients of this expansion can be achieved with the help of a comparatively small number of measurements obtained from spherical t-designs, as shown in "Unique Compact Representation of Magnetic Fields using Truncated Solid Harmonic Expansions". This method was used in "Flexible Selection Field Generation using Iron Core Coil Arrays" to investigate the complex relation between currents of iron core coils and the generated magnetic field. Utilizing a coil sensor, the drive field profile can be measured using the scanners own analog-to-digital converter, shown in "Efficient 3D Drive-Field Characterization for Magnetic Particle Imaging Systems".

A horizontal and vertical field-free line measured inside our Low-Power Iron Magnetic Field Generator.

Publications

[191963]
Title: Current-to-Field Prediction for Non-Linear Magnetic Systems via Neural Networks.
Written by: F. Foerger, P. Jürß, M. Boberg, T. Hau, T. Knopp, and M. Möddel
in: <em>International Journal on Magnetic Particle Imaging</em>. Mar (2025).
Volume: <strong>11</strong>. Number: (1 Suppl 1),
on pages: 1-2
Chapter:
Editor:
Publisher:
Series:
Address:
Edition:
ISBN:
how published:
Organization:
School:
Institution:
Type:
DOI: https://doi.org/10.18416/IJMPI.2025.2503009
URL: https://www.journal.iwmpi.org/index.php/iwmpi/article/view/862
ARXIVID:
PMID:

[www]

Note: inproceedings, magneticfield, ml

Abstract: Accurate magnetic field knowledge is crucial for magnetic particle imaging, affecting performance estimation, sequence generation, and reconstruction. Especially for non-linear field generators, such as those with built-in soft iron, conventional field simulations, such as the finite element method, are computationally demanding. We propose the use of neural networks to predict the coefficients of the spherical harmonic expansions of the fields from the input currents, drastically speeding up current-to-field prediction.