UC Home > Library > UC Research Repository > College of Science > Science: Theses and Dissertations >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10092/7176

Title: A 3D Computer Vision System in Radiotherapy Patient Setup
Authors: Chyou, Te-yu
Keywords: radiotherapy
radiation therapy
patient setup
patient positioning
patient alignment
structured light
Issue Date: 2012
Abstract: An approach to quantitatively determine patient surface contours as part of an augmented reality (AR) system for patient position and posture correction was developed. Quantitative evaluation of the accuracy of patient positioning and posture correction requires the knowledge of coordinates of the patient contour. The system developed uses the surface contours from the planning CT data as the reference surface coordinates. The corresponding reference point cloud is displayed on screen to enable AR assisted patient positioning. A 3D computer vision system using structured light then captures the current 3D surface of the patient. The offset between the acquired surface and the reference surface, representing the desired patient position, is the alignment error. Two codification strategies, spatial encoding, and temporal encoding, were examined. Spatial encoding methods require a single static pattern to work, thus enabling dynamic scenes to be captured. Temporal encoding methods require a set of patterns to be successively projected onto the object, the encoding for each pixel is only complete when the entire series of patterns has been projected. The system was tested on a camera tracking object. The structured light reconstruction was accurate to within ±1 mm, ±1.5 mm, and ±4 mm in x, y, and z-directions (camera optical axis) respectively. The method was integrated into a simplified AR system and a visualization scheme based on z-direction offset was developed. A demonstration of how the final AR-3D vision hybrid system can be used in a clinical situation was given using an anatomical teaching phantom. The system and visualisation worked well and demonstrated the proof of principal of the approach. It was found that the achieved accuracy was not yet sufficient for clinical use. Further work on improving the projector calibration accuracy is required. Both the camera registration process and 3D computer vision using structured light have been shown to be capable of sub-millimeter accuracy on their own. If that level of accuracy can be reproduced in this system, the concept presented can potentially be used in Oncology departments as a cost-effective patient setup guidance system for external beam radiotherapy, used in addition to current laser/portal imaging/cone beam CT based setup procedures.
Publisher: University of Canterbury. Physics and Astronomy
Degree: Master of Science
URI: http://hdl.handle.net/10092/7176
Rights: Copyright Te-yu Chyou
Rights URI: http://library.canterbury.ac.nz/thesis/etheses_copyright.shtml
Appears in Collections:Science: Theses and Dissertations

Files in This Item:

File Description SizeFormat
Chyou_Use_of_thesis_form.pdf101.24 kBAdobe PDFView/Open
thesis_fulltext.pdf7.78 MBAdobe PDFView/Open


Items in UC Research Repository are protected by copyright, with all rights reserved, unless otherwise indicated.


Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback