Abstract:
Aim : The current CT-based radiation therapy planning workflow at Auckland City Hospital requires both CT and MR imaging : CT primarily for attenuation characteristics for radiation dose calculation and MR for soft tissue contrast. To streamline the workflow and to reduce the workload and imaging costs, a MR-only radiation therapy planning workflow is being considered to replace the current CTbased workflow. Methods : An atlas-based method for CT synthesis and segmentation was proposed using a clinically available radiotherapy planning software MIM (MIM Software Inc., Cleveland, OH). A multi-atlas based synthetic CT conversion workflow was developed using MIM's in-built deformable registration tools. MIM offers three types of deformable image registration algorithms (DIR): one based on image intensities (intensity-based), one using contours of anatomical structures of interest (contour-based), and one using a combination of the intensity-based and contour-based methods (hybrid). The conversion workflow was tested on eight prostate cancer, and eighteen head and neck cancer patients. We also tested MIM's atlas-based segmentation to automatically define structure sets required for radiotherapy planning using a single atlas and multi-atlas as a comparison. The automated structures were compared to the ground truth expert-approved structure sets to confirm the clinical accuracy. Results : The intensity-based workflow produced the overall best synthetic CT quality. However, the difference in CT intensities between the synthetic CT and the original planning CT was a lot higher than other research, especially in the head and neck data. Further research is needed to develop a more reliable synthetic CT conversion algorithm. More positively, the segmentation study produced some promising results in both prostate and head and neck data. Better segmentation results were obtained using the multi-atlas than the single atlas. Conclusion : Results suggest that more research needs to be done to create a more reliable CT synthesis and that more consistency in the imaging data is required. Nevertheless, this thesis provides a preliminary work for introducing an MR-only radiotherapy planning workflow.