3d-SMRnet: Achieving a New Quality of MPI System Matrix Recovery by Deep Learning

Magnetic particle imaging (MPI) data is commonly reconstructed using a system matrix acquired in a time-consuming calibration measurement. Compared to model-based reconstruction, the calibration approach has the important advantage that it takes into account both complex particle physics and system imperfections. However, this has the disadvantage that the system matrix has to be re-calibrated each time the scan parameters, the particle types or even the particle environment (e.g. viscosity or temperature) changes. One way to shorten the calibration time is to scan the system matrix at a subset of the spatial positions of the intended field-of-view and use the system matrix recovery. Recent approaches used compressed sensing (CS) and achieved subsampling factors up to 28, which still allowed the reconstruction of MPI images with sufficient quality. In this work we propose a novel framework with a 3d system matrix recovery network and show that it recovers a 3d system matrix with a subsampling factor of 64 in less than a minute and outperforms CS in terms of system matrix quality, reconstructed image quality, and processing time. The advantage of our method is demonstrated by reconstructing open access MPI datasets. Furthermore, it is also shown that the model is capable of recovering system matrices for different particle types.

Subjects

Deep learning

Magnetic particle imaging

Single image super-resolution

System matrix recovering

Options

3d-SMRnet: Achieving a New Quality of MPI System Matrix Recovery by Deep Learning