Line-to-Point Registration with Applications in Geometric Reconstruction of Coronary Stents

Line-to-Point Registration with Applications in Geometric Reconstruction of Coronary Stents By Claire Y. Lin Registration is a process of geometrically transforming one object to correspond to the other. It can utilized to align images, surfaces, or point clouds. The Iterative Closest Point algorithm is widely used in registration to achieve reconstruction of geometric shapes, but has drawbacks such as nondifferentiability. This thesis introduces a line-to-point registration method, useful in imposing a skeletal structure, defined by nodes and edges, on a given set of points in 2D or 3D. This method computes the distance from a point cloud to a skeletal structure using projections, and uses rigid, affine, and nonparametric transformations for distance minimization, taking into account regularization on the nonparametric transformation. The proposed approach can be utilized in the registration of two geometric objects, where one has a known structural skeleton, and the other is a point set. In this thesis, this method is used to achieve correspondence between the undeformed and deformed configurations of a coronary prosthesis, called a bioresorbable stent. The undeformed configuration is represented by a skeleton of the prosthesis based on the manufacturer’s design, and the deformed configuration is represented by a set of points obtained from medical images. Registration is used to automatize the geometric reconstruction of implanted coronary stents in patient-specific cases, to allow Computational Fluid Dynamics (CFD) analysis in the clinical trials at Emory University. Line-to-Point Registration with Applications in Geometric Reconstruction of Coronary Stents