Biomedical Engineering Reference
In-Depth Information
2.8 Validation ...................................................................................................... 34
2.9 Summary and Conclusion .......................................................................... 35
References ............................................................................................... 36
2.1
Introduction
This chapter presents a descriptive account of methods for image registra-
tion. Our intention here is to enable the reader to understand the main con-
cepts behind the different methods without recourse to the underlying
mathematics. All equations have been banished! The mathematics and
details of implementation are left to the following chapter, which has a very
similar structure that will allow the reader to switch between the two for
more detailed descriptions when required.
As stated in Chapter 1, medical image registration has a wide range of
potential applications. These include:
•
Combining information from multiple imaging modalities, for
example, when relating functional information from nuclear medi-
cine images to anatomy delineated in high-resolution MR images.
•
Monitoring changes in size, shape, or image intensity over time
intervals that might range from a few seconds in dynamic perfusion
studies to several months or even years in the study of neuronal
loss in dementia.
•
Relating preoperative images and surgical plans to the physical
reality of the patient in the operating room during image-guided
surgery or in the treatment suite during radiotherapy.
•
Relating an individual's anatomy to a standardized atlas.
To be effective, all these applications require the establishment of spatial
correspondence. What we mean by correspondence in image registration is
explored in this chapter before presenting, in descriptive terms, the various
methods of registration. The process of image registration involves finding
transformations that relate spatial information conveyed in one image to those
in another or in physical space. We relate the type of transformation to the
number of dimensions of the images. We describe the number of parameters,
or “degrees of freedom,” which are needed to describe this transformation
for the different classes of registration algorithm. We introduce the concept of
optimization, in which the computer makes a succession of guesses about the
correct data before converging to an answer that should be close to the correct
one. Issues related to image transformation are discussed, and a few comments
