Bernd Andr Jung1, Simon Bauer1, Michael Markl1
1Dept. of Diagnostic Radiology, Medical Physics, University Hospital, Freiburg, Germany
The aim of this work was to explore how to optimally undersample and reconstruct time-resolved 3D data using GRAPPA. Two different data sets were acquired in a moving phantom with isotropic and anisotropic data matrices. Reconstruction was performed with 3D- (kx,ky,t) and 4D-kernel (kx,ky,kz,t) configurations. For the symmetric data matrix, it was demonstrated that the 4D-kernel configuration leads to better results in terms of error behaviour. However, in a more realistic anisotropic data matrix typically used in clinical applications the different kernel configurations show an opposite behaviour. Furthermore, noise enhancement for 4D-kernel configuration was more pronounced compared to 3D-configurations.