GPU-accelerated parallel image reconstruction strategies for magnetic particle imaging

Published

June 24, 2024

Author(s)

Klaus Natorf Quelhas, Mark-Alexander Henn, Ricardo Farias, Weston L. Tew, Solomon I. Woods

Abstract

Image reconstruction is a fundamental step in Magnetic Particle Imaging (MPI). Since it was developed, several methods have been studied to perform more efficient and accurate reconstructions. One of the challenges of MPI is the fact that the reconstructions are computationally intensive and time-consuming, so choosing the algorithm presents a compromise between accuracy and execution times, which depends on the application. This work addresses this challenge by implementing a number of image reconstruction algorithms to be executed in graphics processing units (GPUs), showing that MPI reconstructions can be accelerated up about 6,000 times, allowing for real-time imaging applications. It also demonstrates how it is possible to combine some of these algorithms in order to obtain high accuracy reconstructions in reduced times. Finally, this work presents the GPU-accelerated calculation of the MPI calibration matrix, which provides the flexibility of redefining scanning and reconstruction parameters during execution.

Citation

Physics in Medicine and Biology

Volume

69

Pub Type

Journals

Download Paper

https://doi.org/10.1088/1361-6560/ad5510

Local Download

Keywords

Magnetic particle imaging, MPI, image reconstruction, parallel computing, GPU, CUDA, Kaczmarz, conjugate gradients, CGNR, simgular value decomposition, SVD

Citation

Natorf Quelhas, K. , Henn, M. , Farias, R. , Tew, W. and Woods, S. (2024), GPU-accelerated parallel image reconstruction strategies for magnetic particle imaging, Physics in Medicine and Biology, [online], https://doi.org/10.1088/1361-6560/ad5510, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=956953 (Accessed October 14, 2025)