The registration, or alignment, of multiple medical scans can provide a plethora of new information about a patient beyond what can be gained from studying the individual scans in isolation. This added information can increase the accuracy in specifying regions of interest, help in radiation therapy planning, and aid in a myriad of other treatment and diagnostic techniques. However, the registration of the scans is very complex, and no panacea exists for adjusting for the numerous variations and distortions that can occur between scans.
This thesis briefly summarizes the current technology of MRI,
CT, PET, and SPECT imaging. Possible uses and benefits of using
registered data sets are then explained in this context. This
is followed by a structured description of the general problem
of data registration and a framework that can be used to solve
it. Current methods that have been utilized by various researchers
are surveyed. The implementation of a surface matching method
and the adaptations and innovations that were utilized is described
in detail. These are analyzed under diverse test conditions involving
both real and simulated data. The surface matching methods that
were developed provide an accurate means for the registration
of rigid body transformations, although many problems remain to
be solved before a robust solution exists.