Parallel MRI: Concepts and Methods - Advance Image Processing in Magnetic Resonance Imaging - page 73

0.4

0.3

0.2

0.1

FIGURE 3.10 Self-calibrated 8-channel head coil-sensitivity maps.

Figure 3.12 shows the reconstructions from four separate algorithms on the

2x-accelerated data. Figure 3.12(a) shows a reconstruction following the SMASH
approach described in the preceding text. As mentioned earlier, the SMASH
reconstruction approach is very sensitive to the coil-sensitivity map estimates. As
this example shows, SMASH reconstruction is not well suited for self-calibrated
sensitivity estimation. A better approach is to use the AUTO-SMASH reconstruc-
tion, shown in Figure 3.12(b). This k-space domain method does not require
estimation of the coil-sensitivity maps and produces a much cleaner image recon-
struction with significant suppression of the aliasing artifacts. As noted earlier,
however, the SMASH and AUTO-SMASH approaches present only an approxi-
mate solution to the parallel MR signal equation. The generalization of these
approaches in GRAPPA provides enough flexibility in reconstruction parameter
selection to effectively suppress the artifacts visible in the other two k-space
methods, as the reconstruction with 8 ACS lines in 3.12(c) illustrates. This
reconstruction is comparable to the SENSE approach, which does minimize the
least-squared error between the acquired data and the signal acquisition model

250

200

150

100

50

0

20 40 60 80 100 120

Encode count

(a) (b)

FIGURE 3.11 Uniform 2x sampling pattern and spatial-domain aliasing result.