LV surface reconstruction from sparse tMRI using Laplacian Surface Deformation and Optimization

Abstract

We propose a novel framework to reconstruct the left ventricle (LV)'s 3D surface from sparse tagged-MRI (tMRI). First we acquire an initial surface mesh from a dense tMRI. Then landmarks are calculated both on contours of a specific new tMRI data and on corresponding slices of the initial mesh. Next, we employ several filters including global deformation, local deformation and remeshing to deform the initial surface mesh to the image data. This step integrates Polar Decomposition, Laplacian Surface Optimization (LSO) and Deformation (LSD) algorithms. The resulting mesh represents the reconstructed surface of the image data. Further more, this high quality surface mesh can be adopted by most deformable models. Using tagging line information, these models can reconstruct LV motion. The experimental results show that compared to Thin Plate Spline (TPS) our algorithm is relatively fast, the shape represents image data better and the triangle quality is more suitable for deformable model.