Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Medical Biophysics

Supervisor(s)

Eugene Wong, Aaron Fenster

Abstract

High-dose-rate brachytherapy (HDR-BT) is an interstitial technique for the treatment of intermediate and high-risk localized prostate cancer that involves placement of a radiation source directly inside the prostate using needles. Dose-escalated whole-gland treatments have led to improvements in survival, and tumour-targeted treatments may offer future improvements in therapeutic ratio. The efficacy of tumour-targeted HDR-BT depends on imaging tools to enable accurate dose delivery to prostate sub-volumes. This thesis is focused on implementing ultrasound tools to improve HDR-BT needle localization accuracy and efficiency, and evaluating dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) for tumour localization.

First, we implemented a device enabling sagittally-reconstructed 3D (SR3D) ultrasound, which provides sub-millimeter resolution in the needle insertion direction. We acquired SR3D and routine clinical images in a cohort of 12 consecutive eligible HDR-BT patients, with a total of 194 needles. The SR3D technique provided needle insertion depth errors within 5 mm for 93\% of needles versus 76\% for the clinical imaging technique, leading to increased precision in dose delivered to the prostate.

Second, we implemented an algorithm to automatically segment multiple HDR-BT needles in a SR3D image. The algorithm was applied to the SR3D images from the first patient cohort, demonstrating mean execution times of 11.0 s per patient and successfully segmenting 82\% of needles within 3 mm.

Third, we augmented SR3D imaging with live-2D sagittal ultrasound for needle tip localization. This combined technique was applied to another cohort of 10 HDR-BT patients, reducing insertion depth errors compared to routine imaging from a range of [-8.1 mm, 7.7 mm] to [-6.2 mm, 5.9 mm].

Finally, we acquired DCE-MRI in 16 patients scheduled to undergo prostatectomy, using either high spatial resolution or high temporal resolution imaging, and compared the images to whole-mount histology. The high spatial resolution images demonstrated improved high-grade cancer classification compared to the high temporal resolution images, with areas under the receiver operating characteristic curve of 0.79 and 0.70, respectively.

In conclusion, we have translated and evaluated specialized imaging tools for HDR-BT which are ready to be tested in a clinical trial investigating tumour-targeted treatment.


Share

COinS