Jean Pierre Bassenge, BCCN Berlin / TU Berlin

Spatiotemporal Undersampling Techniques for Accelerated 4D Flow MR Imaging of the Cerebral Blood Velocity Vector Field

4D flow MR imaging is a promising technique for the diagnostics of abnormalities in cerebral vessel hemodynamics. Nevertheless, long scanning times and insufficient temporal and spatial resolution still hinder it from being used in clinical routine, which motivates the development of fast MR techniques that still preserve imaging quality to a sufficient degree. The goal of this thesis is to explore spatiotemporal undersampling techniques with high acceleration factors (up to ~8-fold net acceleration), and to quantify the arising measurement error in clinically relevant markers such as the maximum velocity of blood flow in each voxel.

To this aim, the performance of three spatiotemporal undersampling schemes, two of which have not previously been employed for 4D flow imaging, was evaluated by retrospective undersampling of fully sampled data sets and subsequent reconstruction based on training data, and compared to standard acceleration methods. This procedure was applied to three different data sets: one in vivo data set acquired at a 3 Tesla MRI scanner, one in vivo data set from the same subject acquired at 7 Tesla, and one phantom measurement using a custom-made flow phantom with an experimental setup for creating pulsatile flow, which was developed as part of this thesis.

The quantification of measurement errors for the different methods suggests the general feasibility of high acceleration factors using spatiotemporal techniques, with consequences for the imaging quality in the same range that standard techniques with a much lower net acceleration factor are known to exhibit. Nevertheless, further research involving more subjects will be required to make those findings more robust.

Additional Information

MSc defence in the international master program Computational Neuroscience of the BCCN Berlin

Organized by

Felix Blankenburg / Robert Martin