Gradient-Echo Pulse Sequences

The basic idea of parallel imaging is to reduce the number of acquisition steps (i.e. phase-encoding steps) by decreasing the sampling density in kspace. Artefacts, in particular aliasing artefacts, that would result from the reduced sampling density are compensated by employing image data from several independent coil elements with different spatial sensitivity profi les for reconstruction. This basic concept of parallel imaging, i.e. reducing the sampling density and increasing the number of coil elements, can be combined with almost all types of MR pulse sequences. The only precondition with respect to the pulse-sequence design is that the phase-encoding sampling density in k-space can be varied, and this is true for virtually all pulse sequences that acquire at least two-dimensional k-space data, i.e. for all sequences apart from exotic techniques such as column-selective MRI or line scan imaging (Maudsley 1980; Gudbjartsson et al. 1996). The latter sequence types can, however, benefi t from some more general parallel-imaging approaches, e.g. by acquiring data from two separate volumes at once. In this case more complex pulse-sequence modifi cations, such as the insertion of special RF pulses, are required (Larkman et al. 2001; Breuer et al. 2005). Parallel imaging can be combined with conventional Cartesian k-space trajectories as well as with non-Cartesian sampling strategies such as radial or spiral trajectories; however, the complexity of the image reconstruction differs substantially depending on the k-space sampling strategy. Parallel imaging with non-Cartesian k-space trajectories requires very complex and time-consuming reconstruction algorithms and, thus, has not yet found general acceptance in clinical routine. More details about parallel imaging with non-Cartesian k-space techniques can be found in Chap. 6. The following sections discuss the application of parallel imaging in several basic pulse sequence types such as spin-echo or gradientecho techniques. Generally, Cartesian k-space sampling is assumed, although many conclusions remain valid for other trajectories as well.