A robotic needle guide for prostate brachytherapy with pre-operative to intra-operative prostate volumes registration

• Main Author: Prananta, Thomas Diego
• Language: English
• Published: University of British Columbia 2010
• Online Access: http://hdl.handle.net/2429/28037

The conventional prostate brachytherapy approach is limited by needle positioning accuracy, needle trajectory option, and prostate motion and deformation between the pre-operative volume study and the seed implant procedure. These limitations increase the risks of post implant complications. In this thesis we develop a robotic needle guide to improve prostate brachytherapy needle placement accuracy and trajectory option as well as a pre-operative to intra-operative prostate volume registration algorithm to address the issue of prostate motion and deformation. Our four degrees of freedom robot provides X-Y axes translational accuracy of 0.12 and 0.1 mm compared to the 5 mm accuracy of the standard needle guide. The robot also provides yaw and pitch angulations with 0.05 degrees accuracy which can be used to reach prostate regions blocked by pubic arch interference. The robot is adaptable to conventional brachytherapy apparatus without adding the clinical procedure time and can be used manually in the case of electronic control failure. The registration approach is based on fitting prostate surfaces into ellipsoids. Pre-operative and intra-operative sagittal view-based volume data are contoured using a novel semi automatic sagittal view-based segmentation algorithm. The resulting contours are fit into ellipsoids whose parameters - centers, orientations, and radii - are used to calculate the registration matrix. The accuracy of the registration algorithm was compared with Optotrak measurement as the gold standard and with the Iterative Closest Point (ICP) algorithm. The result shows that the orientation of the ellipsoid fit is sensitive to user initialization points causing up to 5 mm translational errors and 5.5 degrees angular error. The comparison with ICP shows that the ellipsoid fitting based algorithm is faster but less accurate.