Intensity-based Fluoroscopy and Ultrasound Registration for Prostate Brachytherapy

Thumbnail Image
Karimaghaloo, Zahra
Prostate brachytherapy , Brachytherapy seeds , Transrectal ultrasound , Mutual information , Registration , Prostate cancer , Fluoroscopy
Prostate cancer continues to be the most commonly diagnosed cancer among men. Brachytherapy has emerged as one of the definitive treatment options for early stage prostate cancer which entails permanent implantation of radioactive seeds into the prostate to eradicate the cancer with ionizing radiation. Successful brachytherapy requires the ability to perform dosimetry -which requires seed localization- during the procedure but such function is not available today. If dosimetry could be performed intraoperatively, physicians could implant additional seeds into the under-dosed portions of the prostate while the patient is still on the operating table. This thesis addresses the brachytherapy seed localization problem with introducing intensity based registration between transrectal ultrasound (TRUS) that shows only the prostate and a 3D seed model drawn from fluoroscopy that shows only the implanted seeds. The TRUS images are first filtered and compounded, and then registered to the seed model by using mutual information. A training phantom was implanted with 48 seeds and imaged. Various ultrasound filtering techniques were analyzed. The effect of false positives and false negatives in ultrasound was investigated by randomly masking seeds from the fluoroscopy volume or adding seeds to that in random locations. Furthermore, the effect of sparse and dense ultrasound data was analyzed by running the registration for ultrasound data with different spacing. The registration error remained consistently below clinical threshold and capture range was significantly larger than the initial guess guaranteed by the clinical workflow. This fully automated method provided excellent registration accuracy and robustness in phantom studies and promises to demonstrate clinically adequate performance on human data.
External DOI