In radiotherapy, the treatment is typically pursued by irradiating the patient with high energy x-ray beams conformed to the shape of the tumour from multiple directions. Rather than administering the total dose in one session, the dose is often delivered in twenty to thirty sessions. For each session several settings must be reproduced precisely (treatment setup). These settings include machine setup, such as energy, direction, size and shape of the radiation beams as well as patient setup, such as position and orientation of the patient relative to the beams. An inaccurate setup may result in not only recurrence of the tumour but also medical complications. The aim of the project is to develop a novel image processing system to enable fast and accurate evaluation of patient setup errors in radiotherapy by automatic detection and alignment of anatomical features in images acquired during treatment simulation and treatment delivery. By combining various image processing and mathematical techniques, the thesis presents the successful development of an effective approach which includes detection and separation of collimation features for establishment of image correspondence, region based image alignment based on local mutual information, and application of the least-squares method for exhaustive validation to reject outliers and for estimation of global optimum alignment. A complete software tool was developed and clinical validation was performed using both phantom and real radiotherapy images. For the former, the alignment accuracy is shown to be within 0.06 cm for translation and 1.14 degrees for rotation. More significantly, the translation is within the ±0.1 cm machine setup tolerance and the setup rotation can vary between ±1 degree. For the latter, the alignment was consistently found to be similar or better than those based on manual methods. Therefore, a good basis is formed for consistent, fast and reliable evaluation of patient setup errors in radiotherapy.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:411282 |
| Date | January 2004 |
| Creators | Su, QingLang |
| Publisher | University of Central Lancashire |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://clok.uclan.ac.uk/20692/ |
Page generated in 0.0021 seconds