Construction of 3D realistic purkinje system: A locally linear embedding-based method

The geometry of canine ventricle with purkinje system provides the data basis for the simulation and analysis of the mechanisms of ventricular pathophysiology. The acquisition of the geometry of the purkinje system, however, is very challenging, and traditional construction approaches are mainly based on modeling using fractal geometry. In this paper, we propose a novel locally linear embedding (LLE)-based method to construct 3D anatomical purkinje system of the canine left ventricle (LV). First, we collect the 2D purkinje system data in a canine ventricle and extract the endocardial surface of the canine left ventricle from 3D canine ventricle. Then, LLE is used to map the 3D geometry of the endocardial surface to a 2D subspace, and the 2D purkinje system data is further embedded into this 2D subspace. Finally, LLE is adopted to map both the 3D geometry of the endocardial surface and the 2D purkinje system data back to the 3D space. An experiment is designed to verify the effectiveness of the LLE-based construction method. The proposed method approaches is promising in restoring realistic purkinje system of the left canine ventricle.