Biomedical Engineering Reference
In-Depth Information
A fundamental advantage of probabilistic atlantes is that they allow for the
calculation of statistical probability anatomy maps which can be used to
derive statistical knowledge about the spatial position and variability of ana-
tomical structures.
13.3.3
Analysis of Motion and Deformation Using
Nonrigid Registration
In the case of rigid registration the recovered transformation itself has no clin-
ical significance. It merely expresses the difference between the position and
orientation in two images. However, in the case of nonrigid registration the
recovered transformation may have clinical significance. In particular, the
transformation may be used for the quantification of changes between
images. In these cases the primary goal is not only the transformation which
maps points in one image into their corresponding counterparts in the sec-
ond image, but also the motion and deformation characteristics exhibited by
this transformation.
One such example is the work of Maurer et al.,
57,58
who have used a non-
rigid registration algorithm to detect and quantify intraoperative brain defor-
mation. In this application, pre- and postoperative MR images from patients
undergoing neurosurgery have been acquired using an interventional MR
scanner. After rigid registration of the pre- and postoperative MR images, the
images have been aligned by a nonrigid registration algorithm which maxi-
mizes the normalized mutual information between the pre- and postopera-
tive images. The resulting deformation field can be used to calculate the
volume change throughout the brain on a voxel-by-voxel basis. An example
of this is shown in Figure 13.6.
In a similar effort, Thirion et al.
59
and Rey et al.
47
used a nonrigid registra-
tion algorithm to identify multiple sclerosis lesions and characterize their
change over time. Finally, Thompson et al.
60
have shown that growth patterns
in the developing brain can be studied and analyzed using the deformation
fields obtained by nonrigid registration. From these deformation fields local
rates of tissue dilation, contraction, and shear are calculated.
13.4
Conclusions
Rigid registration techniques have become widely accepted in a variety of
clinical applications. In contrast, nonrigid registration is very much an area
of ongoing research, and most algorithms are still in the stage of develop-
ment and evaluation. One of the main reasons for the successful impact of
rigid registration techniques is the fact that these techniques can be assessed
and validated against a gold standard (see Chapter 6). The lack of a gold stan-
dard for assessing and evaluating the success of nonrigid registration algo-
rithms is one of their most significant drawbacks. Currently, the only
accepted method for assessing nonrigid registration is based on manually
