Automatic segmentation of tumours using Positron Emission Tomography (PET) was recommended for radiotherapy treatment (RT) planning of head and neck (H&N) cancer patients, and investigated in the scientific literature without reaching a consensus on the optimal process. This project aimed at evaluating the performance of PETCbased automatic segmentation (PETCAS) methods and developing an optimal PETC AS process to be used at Velindre Cancer Centre (VCC). For this purpose, ten algorithms were implemented to represent the most promising PETCAS approaches from a systematic review of the literature. The algorithms’ performance was evaluated on filled phantom inserts with variable size, geometry, tumour intensity and image noise. The impact of thick insert plastic walls on both image quantification and segmentation was thoroughly assessed. The PETCAS methods were further applied to realistic H&N tumours, modelled using a printed subresolution sandwich phantom developed and calibrated in house. Results showed that different PETCAS performed best for different types of target objects. An Advanced decision TreeCbased Learning Algorithm for Automatic Segmentation (ATLAAS) was therefore developed and validated for the selection of the optimal PETCAS approach according to the target object characteristics. Finally, a protocol was designed for the use of PETCAS within RT planning at VCC. The protocol was used retrospectively on a group of 10 oropharyngeal cancer patients, and the results highlighted the additional information brought by PET beyond anatomical imaging. In a prospective study on 10 additional patients, PETCAS replaced manual PET/CT delineation, and accounted for up to 33% of the modifications of manually drawn CT/MRI contours to derive the final planning contour. This study demonstrated the usefulness and reliability of the PETCAS method in RT planning, and led to modifying the clinical workflow for H&N patients at VCC. This work has the potential to be extended to other tumour sites and institutions.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:633575 |
| Date | January 2014 |
| Creators | Berthon, Beatrice |
| Publisher | Cardiff University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://orca.cf.ac.uk/69184/ |
Page generated in 0.0028 seconds