Biomedical Engineering Reference
In-Depth Information
is identification of the nature or pattern of change to be quantified. An illus-
trative example is provided by Figure 7.26, which shows a case of an astrocy-
toma that was monitored before and after treatment. An intuitively natural
measurement to make might be the total tumor volume, which might then be
correlated with outcome. However, as shown in the subtraction image,
Figure 7.26c, although the tumor responds to the treatment, that response is
heterogeneous. Thus measurement of a global volume decrease would fail to
capture a critical factor of localized lack of response, which then acts as a
locus for future tumor growth and can result in poor final outcome. The reg-
istered subtraction technique allows the pattern of change to be identified
and can be used to guide measurement as required. This is a powerful feature
of the registration methodology.
7.20
Conclusion
Use of image registration allows changes to be detected on serial examination
in many diseases when conventional approaches produce equivocal or neg-
ative results. This is likely to increase the sensitivity of MRI for many neuro-
logical applications. Effects due to treatment may also be monitored with this
technique. Pediatrics is notable for the fact that there are also changes due to
growth and development, and these may be affected by disease and treat-
ment. The technique appears likely to have many applications.
The work described in this chapter has exclusively employed rigid-body
registration for serial studies of the brain. Rigid-body registration may also
have clinical applications in other organs or body parts, most notably for nor-
mally rigid structures such as bones. In addition to monitoring disease pro-
gression or regression, other emerging applications are perfusion imaging
and dynamic contrast-enhanced angiography. In both these methods, images
are acquired in quick succession following contrast administration, and
uncorrected changes in subject position during the examination can cause
problems. Finally, recent developments in nonrigid registration methods are
likely to have clinical applications when combined with serial MRI examina-
tions. These are outside the scope of this chapter, but various examples of
early work are discussed in Section III.
References
1. Woods, R.P., Mazziotta, J.C., and Cherry, S.R., MRI-PET registration with auto-
mated algorithm,
J. Comput. Assist. Tomogr
, 17, 536, 1993.
2. Friston, K.J., Ashburner, J., Poline, J.B., Frith, C.D., Heather, J.D., and Frackowiak,
R.S., Spatial registration and normalisation of images,
Human Brain Mapping,
2,
16, 1995.
