Abstract
We propose a deformable image registration algorithm that incorporates shape constraints into an intensity-based diffusion algorithm. The shape of a target is represented by a set of landmark pairs delineated by automatic algorithm or by user selected points in an image pair to be registered. The landmark pairs are used as control points to create a spline function in the form of linear combinations of radial bases as shape constraints to regularize the deformation vector field created by the intensity-based algorithm. The algorithm starts with an intensity-based image registration to establish voxel-by-voxel correspondences. For those regions or targets with inaccurate correspondences, a set of landmark control points are placed to perform a shape based registration using radial base spline function. As the radial base spline has limited spatial support, the shape registration has minimal impact outside of the region of interest where the control points are defined. The proposed method is expected to be useful in handling difficult clinical cases in adaptive radiotherapy such as large anatomical changes, non-correspondences, and motion discontinuities, etc. We validate the procedure by assessing the accuracy of structure deformation. Particularly, we have demonstrated the effectiveness of the method in handling several common challenging problems in adaptive therapy: large deformation, non-correspondence and sliding tumor against chest wall. Experimental results with prostate and lung patient cases have shown the potential of the proposed method to be used as practical clinical tool for adaptive re-planning and dose deformation in adaptive therapy.
Keywords: Adaptive radiotherapy, anatomical constraints, automatic segmentation, deformable image registration, intensity – based deformable, splines.
5