Improved seed-based MR and CT image registration for prostate brachytherapy

Abstract

Prostate cancer's high incidence and high survivability motivate its treatment using tightly focused radiation therapy. Brachytherapy treatment, the implantation of radioactive seeds into the prostate, is increasing in popularity, spurred by advances in medical imaging techniques for prostate visualization. Successful brachytherapy requires precise positioning of implant seeds within the pelvic anatomy. Following implantation, precise localization of individual seeds is required to evaluate treatment, but this remains an open challenge. This thesis addresses the seed localization problem with contributions for improving seed-based registration of MR and CT post-implant images. A model for non-rigid, affine prostate motion is presented and demonstrated to improve on current techniques of rigid registration. Also, an evaluation of the benefit of using multiple, rather than a few, seeds is presented, along with a scheme for validating registrations using manually detected seeds in MR and CT volumes. Finally, a scheme for automatic seed-based MR and CT registration by aligning all seeds is suggested, with supporting algorithms for CT seed-finding and unmatched feature registration. A call for an MR seed-finder is issued, for this is the final component needed to achieve automatic and complete seed-based MR and CT registration.