Fast cardiac magnetic resonance imaging using radical trajectories, temporally constrained reconstruction, and graphics processing unit clusters

Cine phase contrast (PC) magnetic resonance imaging (MRI) is a useful imaging technique that allows for the quantitative measurement of in-vivo blood velocities over the cardiac cycle. Velocity information can be used to diagnose and learn more about the mechanisms of cardio-vascular disease. Compared to other velocity measuring techniques, PC MRI provides high-resolution 2D and 3D spatial velocity information. Unfortunately, as with many other MRI techniques, PC MRI su ers from long acquisition times which places constraints on temporal and spatial resolution. This dissertation outlines the use of temporally constrained reconstruction (TCR) of radial PC data in order to signi cantly reduce the acquisition time so that higher temporal and spatial resolutions can be achieved. A golden angle-based acquisition scheme and a novel self-gating method were used in order to allow for exible selection of temporal resolution and to ameliorate the di culties associated with external electrocardiogram (ECG) gating. Finally, image reconstruction times for TCR are signi cantly reduced by implementation on a high-performance computer cluster. The TCR algorithm is executed in parallel across multiple GPUs achieving a 50 second reconstruction time for a very large cardiac perfusion data set.