Applications of deformable image registration: Automatic segmentation and adaptive radiation therapy

The contents of this thesis are best divided into two components: (i) evaluation of atlas-based segmentation and deformable contour propagation and (ii) adaptive radiation therapy using deformable electron density mapping. The first component of this thesis involves the evaluation of two commercial deformable registration systems with respect to automatic segmentation techniques. Overall, the techniques revealed that manual modifications would be required if the structures were to be used for treatment planning. The automatic segmentation methods utilized by both commercial products serve as an excellent starting point for contouring process and also reduce inter- and intra-physician variability when contouring.;In the second component, we developed a framework for dose accumulation adaptive radiation therapy. By registering the planning computed tomography (CT) images to the weekly cone-beam computed tomography (CBCT) images, we were able to produce modified CBCT images which possessed CT Hounsfield units; this was achieved by using deformable image registration. Dose distributions were recalculated onto the modified CBCT images and then compared to the planned dose distributions. Results indicated that deformable electron density mapping is a feasible technique to allow dose distributions to be recalculated on pre-treatment CBCT scans.