Deformable Surface Matching for Statistical Shape Analysis

We propose a deformable surface model for matching anatomically homologous points for a specific region of interest (ROI), based on both surrounding intensity structures and shape characteristics. The initial estimate of the deformable mesh for the individual ROI was derived by transforming the template mesh to the individual image space using nonlinear warping. For the deformable surface model, we defined an external energy term with a distance transformation of the image and internal energy terms with the stretch and smoothness of the mesh. We emphasized energy minimization at the nodes having high surface curvature of the template since curvatures characterize the surface shape. During energy minimization process, weights of the energy components at each vertex were updated in proportion to the energy difference between the template surface and the deformable surface. Our proposed surfacefeature weighted and adaptive weight-updating surface model was then applied to the analysis of the caudate nucleus.