GPU accelerated multi-patch reconstruction

The multi-patch approach in magnetic particle imaging is used to capture large field of views. System-matrix-based image reconstruction for this approach often considers a joint system of equations to minimize artifacts. Due to the prohibitive size of this inverse problem, reconstructions rely on iterative algorithms that do not need to keep the entire system matrix in memory. This work shows a graphical processing unit accelerated implementation of a generalized multi-patch operator. The achieved runtime improvements allow for multi-patch reconstructions using different optimization algorithms, which in turn allow for a flexible choice of regularization terms.

DDC Class

621.3: Electrical Engineering, Electronic Engineering

004: Computer Sciences

681.2: Testing, Measuring, Sensing Instruments

Publication version

publishedVersion

Lizenz

https://creativecommons.org/licenses/by/4.0/

Name

816.pdf

Type

Main Article

Size

769.15 KB

Format

Adobe PDF

Options

GPU accelerated multi-patch reconstruction