Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Biomedical Engineering

Supervisor

Dr. Aaron Fenster

2nd Supervisor

Dr. Aaron Ward

Joint Supervisor

3rd Supervisor

Dr. Jagath Samarabandu

Joint Supervisor

Abstract

Prostate biopsy is the clinical standard for cancer diagnosis and is typically performed under two-dimensional (2D) transrectal ultrasound (TRUS) for needle guidance. Unfortunately, most early stage prostate cancers are not visible on ultrasound and the procedure suffers from high false negative rates due to the lack of visible targets. Fusion of pre-biopsy MRI to 3D TRUS for targeted biopsy could improve cancer detection rates and volume of tumor sampled. In MRI-TRUS fusion biopsy systems, patient or prostate motion during the procedure causes misalignments in the MR targets mapped to the live 2D TRUS images, limiting the targeting accuracy of the biopsy system.

In order to sample smallest clinically significant tumours of 0.5 cm3with 95% confidence, the root mean square (RMS) error of the biopsy system needs to be

The target misalignments due to intermittent prostate motion during the procedure can be compensated by registering the live 2D TRUS images acquired during the biopsy procedure to the pre-acquired baseline 3D TRUS image. The registration must be performed both accurately and quickly in order to be useful during the clinical procedure. We developed an intensity-based 2D-3D rigid registration algorithm and validated it by calculating the target registration error (TRE) using manually identified fiducials within the prostate. We discuss two different approaches that can be used to improve the robustness of this registration to meet the clinical requirements. Firstly, we evaluated the impact of intra-procedural 3D TRUS imaging on motion compensation accuracy since the limited anatomical context available in live 2D TRUS images could limit the robustness of the 2D-3D registration. The results indicated that TRE improved when intra-procedural 3D TRUS images were used in registration, with larger improvements in the base and apex regions as compared with the mid-gland region. Secondly, we developed and evaluated a registration algorithm whose optimization is based on learned prostate motion characteristics. Compared to our initial approach, the updated optimization improved the robustness during 2D-3D registration by reducing the number of registrations with a TRE > 5 mm from 9.2% to 1.2% with an overall RMS TRE of 2.3 mm.

The methods developed in this work were intended to improve the needle targeting accuracy of 3D TRUS-guided biopsy systems. The successful integration of the techniques into current 3D TRUS-guided systems could improve the overall cancer detection rate during the biopsy and help to achieve earlier diagnosis and fewer repeat biopsy procedures in prostate cancer diagnosis.

Share

COinS