Conference Proceedings
| [191960] |
| Title: Denoising the system matrix with deep neural networks for better MPI reconstructions. |
| Written by: A. Tsanda, K. Scheffler, S. Reiss, and T. Knopp |
| in: <em>International Journal on Magnetic Particle Imaging</em>. Mar (2025). |
| Volume: <strong>11</strong>. Number: (1 Suppl 1), |
| on pages: |
| Chapter: |
| Editor: |
| Publisher: |
| Series: |
| Address: |
| Edition: |
| ISBN: |
| how published: |
| Organization: |
| School: |
| Institution: |
| Type: |
| DOI: 10.18416/IJMPI.2025.2503047 |
| URL: https://www.journal.iwmpi.org/index.php/iwmpi/article/view/810 |
| ARXIVID: |
| PMID: |
Note: inproceedings, ml
Abstract: Magnetic Particle Imaging commonly relies on the system matrix (SM) to reconstruct particle distributions, but noise during acquisition limits both its resolution and image quality. Traditionally, noise reduction requires averaging multiple measurements, which increases acquisition time. This paper presents a deep neural network trained on simulated SMs and measured background noise, which effectively generalizes to real-world data. The model recovers higher frequency components of the SM and serves as a general pre-processing step, enhancing image reconstruction quality while reducing the need for extensive averaging, thus accelerating SM acquisition.
| [191960] |
| Title: Denoising the system matrix with deep neural networks for better MPI reconstructions. |
| Written by: A. Tsanda, K. Scheffler, S. Reiss, and T. Knopp |
| in: <em>International Journal on Magnetic Particle Imaging</em>. Mar (2025). |
| Volume: <strong>11</strong>. Number: (1 Suppl 1), |
| on pages: |
| Chapter: |
| Editor: |
| Publisher: |
| Series: |
| Address: |
| Edition: |
| ISBN: |
| how published: |
| Organization: |
| School: |
| Institution: |
| Type: |
| DOI: 10.18416/IJMPI.2025.2503047 |
| URL: https://www.journal.iwmpi.org/index.php/iwmpi/article/view/810 |
| ARXIVID: |
| PMID: |
Note: inproceedings, ml
Abstract: Magnetic Particle Imaging commonly relies on the system matrix (SM) to reconstruct particle distributions, but noise during acquisition limits both its resolution and image quality. Traditionally, noise reduction requires averaging multiple measurements, which increases acquisition time. This paper presents a deep neural network trained on simulated SMs and measured background noise, which effectively generalizes to real-world data. The model recovers higher frequency components of the SM and serves as a general pre-processing step, enhancing image reconstruction quality while reducing the need for extensive averaging, thus accelerating SM acquisition.
