Progressive Surface Reconstruction of Cardiac Anatomy from Sparse and Scattered Points for Heart Mapping Procedure

Abstract

In computer aided radiofrequency catheter ablation of the heart, intraoperative three-dimensional reconstruction of the intracardiac cavity is a very important step. In this paper, we propose a point-by-point surface reconstruction algorithm based on scattered points and guarantee the properties of local reconstruction and surface smoothing. The algorithm firstly builds a generic self-adapted grid model to enclose and approximate to the initial sampled points (at least four points) captured from endocardial surface. Then, the grid model is deformed when adding a newly sampled point. To avoid the linear increase of computation with the increasing of the number of points and unbalanced neighborhood, we propose a local region search criterion based on orthogonal least squares and a neighborhood partition method. Such criterion can guarantee selecting the minimum local balanced points to complete each step of grid deformation. Four normal and six abnormal cardiac computed tomography angiography models specific to the left atrium are used to verify our method. The results show that the surface reconstruction errors as well as the reconstruction time per update are smaller than the state-of-theart methods applied to progressive reconstruction using sparse and scattered points. Furthermore, a cardiac phantom of right ventricle is used to verify the 3D surface reconstruction by our method. The experimental results confirm the feasibility of the proposed algorithm in practical operation.