Biomedical Engineering Reference
In-Depth Information
Chapter 10
Cross-Entropy, Reversed Cross-Entropy, and
Symmetric Divergence Similarity Measures for
3D Image Registration: A Comparative Study
Yang-Ming Zhu
1
and Steven M. Cochoff
1
10.1
Introduction
Image registration is a process that quantitatively relates the information in
one image to that in another image by determining a one-to-one transformation
between coordinates in the two image spaces. Medical image registration is be-
coming increasingly useful in research and patient care (see [1, 2, 3]). Different
imaging modalities often times provide unique and complementary informa-
tion. Multimodality image registration makes it possible to combine structural
(computed tomography or magnetic resonance images) and functional (positron
emission tomography or single photon emission tomography) information to
improve diagnostic accuracy and aid surgical and/or radiotherapeutic planning.
Registration of the same modality images acquired at different times allows
clinicians to assess lesion progression/regression or treatment effectiveness. In
an interactive, image-guided surgery environment, registration of preoperative
images with the physical space is an overriding requirement.
Many algorithms have been employed to register medical images and have
recently been reviewed and classified (see [4, 5]). Earlier work prior to 1993 has
also been reviewed (see [6, 7]). Fitzpatrick and his colleagues evaluated 16 of
