Reconstruction of the magnetic particle imaging system matrix using symmetries and compressed sensing (Q277664)

The authors discuss different ways to reduce the number of calibration scans necessary to reconstruct the magnetic particle imaging (MPI) system matrix, in which each row describes the spatial distribution of the corresponding frequency component, while each column describes each calibration scan. After the symmetry method and the compressed sensing method being introduced, two different possibilities for combining symmetry properties with compressed sensing are considered in the reconstruction of the MPI system matrix. Next, the authors calculate the deviation from the fully sampled system matrix. It is checked, among other comparisons, if the combination of mirroring with compressed sensing works equally well to the different mirror combinations. After analyzing the normalized root mean squared error of the MPI system matrix for the different reconstruction techniques, it is studied how the error translates to the MPI phantom images when using the approximated system matrices for image reconstruction. The experiments in this work have been carried out on 2D measurement data.