Parallel MRI: Concepts and Methods - Advance Image Processing in Magnetic Resonance Imaging - page 77

than the GRAPPA reconstruction, the resolution of the boxes on the right is
maintained. Furthermore, the inherent ability within SPACE RIP to modify the
phase encode selection allows one to choose different phase encode lines that
distribute the residual artifact more evenly over the field of view while maintaining

the required resolution. This is illustrated in Figure 3.16(c) , using the 4x-expo-
nentially-weighted sub-sampling distribution given in Figure 3.15 .

The cost of this flexibility in sub-sampling is computational complexity.
SPACE RIP reconstructions typically take six to ten times longer to compute than
GRAPPA. This is because for each image, SPACE RIP must solve a system of
equations for each column, with each system of size LP-by-N, where L is the
number of coils, P is the number of phase encodes acquired, and N is the size
of the FOV in the phase-encode direction. In contrast, GRAPPA needs to solve
a system of equations for each missing line of k-space: L coils M-P lines (size

of FOV-number of phase encodes). For each of these systems the matrix equation
that solves the GRAPPA parameter fit equation is of size p-by-N, where p is the
number of parameters to determine and N is the size of the FOV along the readout
direction. With a smaller system matrix size, GRAPPA requires significantly less
computational resources to compute.
The ability to provide reconstruction of irregularly sampled data with minimal
computational resources has led to rapid adoption of GRAPPA in clinical settings.
However, with rapidly improving computational resources, it is anticipated that
SPACE RIP will become more prevalent, given its greater flexibility in tailoring
sub-sampling strategies for both artifact suppression and resolution improvement.

×

3.4.3 E XAMPLE 3: I N V IVO A PPLICATIONS

Parallel imaging has found wide applicability in clinical imaging, because it can
be used to improve both spatial and temporal resolution. Our final examples
illustrate two applications in which parallel MR imaging has found strong
popularity and a promising future. In cardiac imaging, shown in Figure 3.17,
the standard clinical examination currently requires a patient breath hold of

FIGURE 3.17 Axial cardiac image.