Image registration in adaptive radiation therapy

This thesis focuses on the development, validation and application of image registration software in adaptive radiation therapy. The contents can be divided into three components: deformable image registration validation, rigid registration based daily patient positioning and deformable image registration for dose accumulation. In the first component, a protocol was developed for the validation of commercial deformable registration systems, applicable to a wide range of applications. The protocol was used to assess the capabilities of a commercial system and results suggest that deformable registration could potentially be optimized by treating applications separately instead of using one algorithm for all applications.;The third component of this thesis involves the use of deformable registration and dose accumulation in the comparison of treatment and planning doses. Using correlation coefficient optimization and assuming B-spline parameterized deformations, we demonstrated that deformable registration can be improved by performing separate registrations over each clinically relevant region on interest. Our deformable model was incorporated into a dose accumulation framework and cumulative treatment doses were compared to those that were planned. In addition, cumulative doses that would have been delivered had patient positioning been based on bone matching and prostate matching were also evaluated. These studies may suggest that daily anatomical variations play a greater role in treatment dosimetry than does the selection of registration based image guidance procedure.;The use of rigid registration for daily positioning of helical tomotherapy prostate patients was investigated in the second component. We quantified alignment discrepancies between daily treatment MVCT images and their corresponding planning CT images resulting from different automated rigid registration schemes. Based on alignments, errors in prostate positioning that would occur if patient repositioning was based on mutual information optimization of entire images or simply bony anatomy were evaluated, with the latter having a 20% decreased average prostate misplacement. The dosimetric implications of performing patient positioning based on either bony anatomy matching or prostate matching in treatment and planning images were also investigated. Prostate doses were fairly insensitive, however, doses to the radiation sensitive bladder and rectum varied with not only positioning strategy, but also the direction of daily prostate motion.