Key Technologies Of Image Registration For Accurate Radiotherapy

As one of the key technologies in accurate radiotherapy, medical image registration was the basis of disease diagnosis, target delineation, accurate positioning and accurate treatment delivery. Multi-modality rigid image registration, such as CT/MRI and CT/PET, integrated more useful image information for physicians; and deformation registration can provide the changes in the anatomical structure of the patients. However, the automatic registration in the clinical application cann’t be achieved only by markers based rigid registration method; in addition, deformable registration method only reflected the local deformation in the process of radiotherapy, which was difficult to meet the clinical requirements of low contrast or artifact images. In order to meet the requirements of image registration during the treatment planning, patient positioning as well as delivery process, related study has been done in this thesis on the key issues of medical image registration for the procedure of radiotherapy.First, a rigid registration method based on local weighted mutual information was developed according to image intensity and spatial information, which effectively solved the local extreme problem in mutual information based registraion. The experimental results showed that sub-milimeter accuracy was achieved, and can be applied in the clinical practice to help physicans to diagnose the disease and delineate the target volume.Secondly, a deformable CT/CT image registration method based on contours information was proposed, which adopting the contours of the planning CT and the treatment CT as prior knowledge to realize the registration of the soft tissue with low contrast. The experiment results indicated that the contours guided Demons method could be used to deal with the images with low contrast and high noises, the average NCC increased to 0.98±0.01, and the deformable field could propogate the contours of the planning CT images to that of the treatment CT images, thus greatly reducing the workload of physicians. Thirdly, a deformable image registration method based on intentsity correction was proposed to deal with CT and cone beam CT image registration. An intensity correction step was incorporated into each iteration of Demons deformable force, during which the intensity of cone beam CT was locally corrected according to that of CT images. Then CT and cone beam CT image registration was therefore treated as mono-modality image registration. The experimental results showed that the deformed CT images was severely distorted due to the serious artifacts of cone beam CT; on contrary, there was little distortion of the registrated CT images by the intensity corrected Demons method and the average NCC value was increased to 0.95±0.03, which can provide good technical support for monitoring the deformation changes of tumors and the surrounding organs in real time during the radiotherapy treatment.