Research Of Deformable Image Registration For Adaptive Radiation Therapy

Radiation therapy is one of the key techniques for cancer treatements, and above50%of the cancer patients need to receive radiation therapy. Current radiation therapy techniques, such as3-Dimensional conformable radiation therapy (3D-CRT) and intensity modulated radiation therapy (IMRT), have been able to delivere accurate dose to the hypothetical static target. However, due to the complexity of treatment delivery and variation inpatient/tumor intrafraction and interfraction position, treatment may still pose risks for a target miss. For instance the patient boby is not static but keep changing during the entire treatment process, tumor may shrink as the treatment proceed and the filling and shape of the bladder and rectum keep changing from day to day. Adaptive Radiation Therapy (ART) is proposed to solve this problem. ART allows real-time treatment adaptations based on the current patient anatomy and geometry. In a typical online ART process, computed tomography (CT) image is usually acquiredprior to the treatment course for treatment planning purposes. Before each treatment fraction, a cone-beam computed tomography(CBCT) image (or CT image) is then obtained, on which the treatment plan is redesigned to account for setup errors, deformations oftumor and other organs, as well as the change of their relative locations. Deformable image registration (DIR) technique plays an important role in this process to establish a correspondence between voxels in the CT and the CBCT/CT for various purposes, for instance, transferring the organ contours from the planningCT images to the daily CBCT/CT images. It is hence desirable to have an accurate and robustDIR algorithm to facilitate this step.The CT-CBCT DIR or CT-CT DIR is still an open problem because current DIR algorithms are not capable of handling all the DIR problems between CT and CT/CBCT. If CBCT image is used as the static image, CT-CBCT DIR is considered to be an inter-modality DIR problem:Although CT and CBCT are reconstructed under the same physical principles, the intensity (Hounsfield Units (HU)) consistency between CT and CBCT images is violated due to many reasons. Firstly, almost all current commercial systems reconstruct CBCT images using the well-known FDK algorithm. As a fundamental limitation of this algorithm, CBCT quality degrades with increasing of cone angle. Second, scatter contamination also leads to severe cupping, streak artifacts and degrade of image contrast. In a typical clinical CBCT system for radiotherapy scatter-to-primary ratio (SPR) may even exceed100%, although many methods have been proposed to correct scatter artifacts, it is still an open problem. Third, the gantry mounted bowtie filter in CBCT system may wobble, as the gantry rotates, which can result in crescent artifacts. Moreover, there are also other factors that contribute to the intensity inconsistency between CT and CBCT, e.g. different level of noise, beam hardening effects and motion.Another challenge for DIR between CT and CBCT images is the CBCT truncation problem, which is common in ART due to the following several reasons. Firstly, limited size of the detector of a linear accelerator (linac)-integrated CBCT yields a field of view (FOV) roughly27cm (full fan) or48cm (half fan) in diameter, which is much smaller than the FOV in a conventional CT. Secondly, the patient is usually positioned such that the tumor is around the isocenter of the linac. If the tumor is located far from the body center, part of the patient body may be outside the FOV. Thirdly, in order to reduce the imaging dose to a patient, it is sometimes preferable to further restrict the CBCT FOV to the volume of interest (VOI) sufficient for the positioning purpose, as recommended by AAPM Task Group75. This can be achieved by collimating down the CBCT fan angle, leading to a truncation in the CBCT image.While using the CT image as the static image, current DIR algorithms might also fail in some scenarios. For instance, many gynecologic cancer patients receive both IMRT and high-dose-rate (HDR) brachytherapy, during which an applicator is inserted into the vagina and hence appears in the CT image. DIR beween the HDR CT and IMRT CT image is necessary to map one of the images to another for dose accumulation. Yet, DIR between the HDR CT image with applicator and the IMRT CT image without applicator violates the fundamental assumption of most DIR algorithms that there should be point-to-point correspondence in the two images to be registered. Therfore, common DIR algorithms may introduce severe DIR errors if applied directly to the DIR of HDR CT and IMRT CT images.This research focused on solving serveral critical problems of the deformable image registration in ART, and had proposed a series of Demons model based improved DIR algorithms. This research mainly contains:(1) To propose an improved demons-based deformable registration algorithm by adding additional demons force and reallocating the bilateral forces to accelerate convergent speed. A novel energy function is proposed as similarity measure, and BFGS method is utilized for optimization to avoid specifying the numbers of iteration. Mathematical transformed deformable CT images and home-made deformable phantom were used to validate the accuracy of our improved algorithm, and the effectiveness of applying it for contour recontouring is tested as well. Results show that the improved algorithm has relative high registration accuracy and speed when compared with the classic demons algorithm and optical flow based method.(2) We propose and evaluate a modified demons algorithm embedded with a simultaneous intensity correction step, called Deformation with Intensity Simultaneously Corrected (DISC). Rather than estimating a global mathematical transformation model between CT and CBCT images, our method corrects CBCT intensity of each voxel at every iteration step of demons by matching the first and the second moments of the voxel intensities inside a patch around this voxel with those in the CT image. Quantitative evaluations of our method are performed by using both a simulation data set and six clinical head-and-neck cancer patient data sets. It is found that DISC can handle CBCT artifacts and the intensity inconsistency issue and therefore improves the registration accuracy when compared with the original demons. (3) Truncation of a cone-beam computed tomography (CBCT) image, mainly caused by the limited field of view (FOV) of CBCT imaging, poses challenges to the problem of deformable image registration (DIR) between CT and CBCT images in adaptive radiation therapy (ART). The missing information outside the CBCT FOV usually causes incorrect deformations when a conventional DIR algorithm is utilized, which may introduce significant errors in subsequent operations such as dose calculation. In this research, we propose to solve this problem by employing a hybrid deformation/reconstruction algorithm. As opposed to deforming the CT image to match the truncated CBCT image, the CT image is deformed such that its projections match all the corresponding ones for the CBCT image. An iterative forward-backward projection algorithm is developed. Five simulated head-and-neck cancer patient cases and a real truncation data set are used to evaluate our algorithm. It is found that our method can accurately register the CT image to the truncated CBCT image, and it is robust against image truncation.(4) Deformable registration of brachytherapy CT images and external beam CT images is challenging due to the presence of the applicator in the brachytherapy image. We have developed a method to adapt DIR algorithms to this situation. A segmentation step is used to remove the applicator, which is imaged in the HDR CT image but not in the IMRT image, leaving an empty virtual cavity. An artificial deformation vector field is then generated by solving the Navier-Stokes equationin the tissue around the applicator to deflate this cavity toward the center at each slice and restore the vagina to a state without applicator. Different schemes are employed for two different registration tasks, i.e., deforming HDR CT image to IMRT CT image, or vice versa. In both cases, demons algorithm is used as the main DIR method to perform registration. The generated final deformation vector fields are used to deform the dose grid from one CT anatomy to another for dose accumulation.