Analysis of Dynamic Magnetic Resonance Elastography Data - Advance Image Processing in Magnetic Resonance Imaging - page 456

FIGURE 14.3 Left: Original band-pass-filtered wave image (for one of eight phase
offsets). Center: Data after top-down directional filtering. Right: Data after bottom-up
directional filtering (contrast increased by 5X).
boundaries of the phantom, giving rise to interference patterns. The effect of
selecting waves propagating in the top-down and bottom-up directions using the
direction-filtering approach described earlier is shown in the center and right
panels. The simplification of the wave field is further evidenced by the amplitude
maps in Figure 14.4, which show the amplitude of the harmonic motion at the
driving frequency at each voxel (from the first positive temporal frequency plane,
extracted by the Fourier transform of the phase offsets, as described earlier). The
amplitude map for the nondirectionally filtered data (left) contains many nodes,
i.e., areas of low motion due to cancellation between the main top-down wave
and reflections from the inclusions and the bottom wall. Again, the top-down
filtered data is far smoother, and the nodes evident in the original wave field are
greatly reduced (right). The result of the four inversion algorithms described in
the preceding text are shown in Figure 14.5 . Profiles through the reconstructions
are shown in Figure 14.6 . All four algorithms clearly show the four inclusions,
with differing levels of artifacts in the lower regions of the phantom. The reference
FIGURE 14.4 The amplitude of the harmonic motion at the driving frequency for the
band-pass-filtered (left) and the top-down direction-filtered (right) data sets.