Biomedical Engineering Reference
In-Depth Information
FIGURE 13.4
A maximum intensity projection (MIP) of a contrast-enhanced MR mammography: (a) without
registration and (b) with nonrigid registration.
22
with rigid and affine registration techniques has demonstrated the superior
performance of the nonrigid registration technique for contrast-enhanced
MR mammography.
45
For a number of other registration tasks it is necessary to model rigid as well
as nonrigid deformations. For example, in the registration of images of the
spine, the vertebrae of the spine are rigid and do not deform, while the sur-
rounding tissue can deform in a nonrigid fashion. Little et al.
18
have recently
shown how the constraints of rigid bodies can be incorporated into a spline
based transformation using radial basis functions. This results in an interpo-
lating solution that is a summation of a linear term corresponding to the rigid
bodies, and a basis function which smoothly tends to zero at the surface of the
rigid bodies. The resulting transformation is exact at rigid bodies, given the
rigid body transformation, and provides smooth interpolation elsewhere.
Other reasons for tissue deformation may be changes over time such as
tumor growth or tissue deformation due to external forces. For example, in
many image-guided surgery applications it is necessary to align preoperatively
acquired images with intra- or postoperative images. To model soft tissue
deformation in a physically more plausible way, Edwards et al.
38
developed a
three-component model, described in Section 13.2.5. This model incorporates
the different deformation characteristics of rigid, elastic, and fluid structures,
and has been tested on CT and MR slices of the brain acquired before and after
surgery for placement of electrode mats on the brain surface prior to excision of
areas of focal activity in the treatment of epilepsy.
Other applications of nonrigid registration for intrasubject registration
include the monitoring of temporal changes in serial MR images
46,47
and
the analysis of rest and stress cardiac SPECT images.
23
