International Journal on Magnetic Particle Imaging IJMPI
Vol. 12 No. 1 Suppl 1 (2026): Int J Mag Part Imag
https://doi.org/10.18416/IJMPI.2026.2603019

Proceedings Articles, ID 959

Unsupervised deep-learning approach for time domain signal denoising in MPI

Main Article Content

Felix Dahms (Chair of Imaging and Computer Vision, RWTH Aachen University, Aachen, Germany), Volkmar Schulz (1) Chair of Imaging and Computer Vision, RWTH Aachen University, Aachen, Germany; 2) Fraunhofer Institute for Digital Medicine MEVIS, Forckenbeckstrasse 55, Aachen, Germany), Franziska Schrank (1) Chair of Imaging and Computer Vision, RWTH Aachen University, Aachen, Germany; 2) Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany)

Copyright (c) 2026 Felix Dahms, Volkmar Schulz, Franziska Schrank

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract


Denoising is crucial in Magnetic Particle Imaging (MPI) since the measured signals are inherently affected by noise and distortions that degrade the reconstruction quality. This work explores an unsupervised deep-learning approach for time-domain denoising of real-world MPI data, which is based on a Wasserstein Generative Adversarial Network (WGAN). The results show that unsupervised deep learning-based denoising can enhance the signal quality of 1D MPI data by increasing the sensitivity and reducing reconstruction artifacts.

Article Details

References

[1] S. Dittmer, C.-B. Schönlieb, and P. Maass. Ground truth free denoising by optimal transport. Numerical Algebra, Control and Optimization, pp. 34–58, 2024, doi:10.3934/naco.2022017.
[2] T. Miyato, T. Kataoka, M. Koyama, and Y. Yoshida, Spectral normalization for generative adversarial networks, in International Conference on Learning Representations, 2018.