.png){kind=icon}
.png){kind=icon}
Return
Higher order regularization using harmonic eigenfunctions for model-based reconstruction in magnetic particle imaging
.png){kind=icon}
PRE
AI
DOI：10.1088/1361-6420/ae3b68.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Magnetic particle imaging (MPI) is a recent imaging modality where superparamagnetic nanoparticles are employed as tracers. The reconstruction task is to obtain the spatial particle distribution from a voltage signal induced by the particles. Generally, in computational imaging variational reconstruction techniques are common and rely on a mathematical model to describe the underlying physics. For the MPI reconstruction task we propose a model-based variational reconstruction technique which incorporates a higher order regularizer, where the regularizer is diagonalized by harmonic eigenfunctions. The proposed image reconstruction algorithm features two major stages: in the first stage, the core stage, the components of the MPI core response are reconstructed. This is the MPI-specific data approximation task which we formulate as a variational problem incorporating the higher order regularizer. The relationship between the particle distribution, the MPI core response and the measured data is given by a mathematical model which was introduced in our earlier research. According to this model the MPI core response is tied to the particle distribution by convolution. Therefore the outcome of the core stage yields the data for the second stage, the deconvolution stage, in which the final reconstructed image is produced by solving an ill-posed deconvolution problem in a robust way relying on earlier research. Interestingly, the quality of the final image depends significantly on the quality of the result of the core stage. A contribution is thus the enhancement of the core stage via higher order regularization. We provide a theoretical foundation for our approach and demonstrate its benefit with numerical examples.
Keywords:
image reconstruction
data approximation
energy minimization
regularization
Laplacian
bi-Laplacian
harmonic eigenfunctions