Advanced Image Reconstruction Methods in MRI - Advance Image Processing in Magnetic Resonance Imaging - page 52

(a)

(b)

FIGURE 2.4 Dynamic images of a chest tumor at two time points after injection of a
contrast agent: (a) SENSE reconstruction ( R = 3, L = 3), and (b) improved SENSE recon-
struction by the proposed method.

ρ r

for each data frame, from which was derived using the GS model. As can be
seen, the regularized SENSE reconstruction (Figure 2.4b) is significantly better
than that from the standard SENSE algorithm (Figure 2.4a).

2.5 CONCLUSION

Image reconstruction from limited Fourier data is a classical problem in tomo-
graphic imaging. Although a general solution to this problem is not available, a
number of practical techniques have emerged, which can provide optimal (or
close-to-optimal) solutions to a particular application problem, leading to signif-
icant improvements in image quality. This chapter provided a tutorial discussion
of some representative techniques, including parametric and nonparametric meth-
ods for superresolution image reconstruction from limited Fourier data and reg-
ularization methods for image reconstruction from multichannel undersampled
Fourier data. The chapter is also intended to provide some basic background
knowledge of the area for the reader to apply these techniques to particular
problems or to further improve them.