Many applications require the adaption of acquisition to motion of the patient. Beside active motion, which can never be avoided completely, additional respiratory and cardiac motion has to be considered. In order to minimize the impact of motion on the measurements, MR acquisitions are commonly adapted to motion using physiological signals of the patient. Thus, an optimal use of the restricted time slots available for acquisition is required.

Acquisition time can be reduced by skipping some measurement data. However, using conventional methods, this leads to artifacts in the reconstructed image. Iterative reconstruction techniques promise to reduce artifacts in the reconstruction of undersampled measurement data. For an application in clinical routine, an acceptable reconstruction time is mandatory. Therefore, an important feature of this work is the acceleration of the algorithms by leveraging GPUs for compute-intensive tasks.