Biomedical Engineering Reference
In-Depth Information
images are blurred prior to registration (high spatial frequency compo-
nents are removed), the interpolation errors are smaller, so errors in the
registration are less. Although the loss of resolution that results from blur-
ring is a disadvantage, registration errors caused by interpolation errors
can be greater than the loss of precision resulting from blurring.
3.5.3
Transformation for Intermodality Image Registration
It should be emphasized that these interpolation issues are more critical for
intramodality registration, where accuracy of considerably better than a
voxel is frequently desired, than for intermodality image registration. In
intermodality registration, one image is frequently of substantially lower
resolution than the other, and the desired accuracy is of the order of a single
voxel at the higher resolution. Furthermore, it is common for the final regis-
tration solution to be used to transform the lower resolution image to the
sample spacing of the higher resolution modality. Interpolation errors are
still likely to be present if trilinear interpolation is used without care, and
may slightly reduce the registration accuracy or degrade the quality of the
transformed images.
3.6
Conclusions
In this chapter, notation for the image registration problem has been intro-
duced, emphasizing the importance of change in image overlap and image
resampling in the registration problem. Various image registration algo-
rithms based on corresponding features or image intensity values were then
described. Until recently, the great majority of image registration algorithms
was restricted to rigid-body or affine transformations, and the algorithms
described here reflect that emphasis. Recently, nonaffine registration to compen-
sate for tissue deformation or differences between subjects has become an area
of active research. Many of the similarity measures described in Section 3.4.3 can
also be applied to nonaffine registration problems by increasing the number of
dimensions in the search space. A more thorough treatment of nonaffine regis-
tration is given in Chapter 13.
For image-to-physical registration, points and surfaces are widely used for
registration because these can easily be identified on the patient in the oper-
ating room using a tracked localizer system (discussed in more detail in
Chapter 12). For image-to-image registration, the great majority of registra-
tion algorithms use intensity information. The most generally applicable of
these algorithms are currently based on information theory.
One appeal of these information theoretic approaches, apart from their suc-
cess, is the mystique that surrounds the word entropy. An interesting anec-
dote to emphasize this point comes from a conversation between Shannon
