From medical images to anatomically accurate finite element grids

The successful application of computational modelling of blood 1ow for the planning of surgical and interventional procedures to treat cardiovascular diseases strongly depends on the rapid construction of anatomical models. The large individual variability of the human vasculature and the strong dependence of blood 1ow characteristics on the vessel geometry require modelling on a patient-speci*c basis. Various image processing and geometrical modelling techniques are integrated for the rapid construction of geometrical surface models of arteries starting from medical images. These discretely de*ned surfaces are then used to generate anatomically accurate *nite element grids for hemodynamic simulations. The proposed methodology operates directly in 3D and consists of three stages. In the *rst stage, the images are *ltered to reduce noise and segmented using a region-growing algorithm in order to obtain a properly de*ned boundary of the arterial lumen walls. In the second stage, a surface triangulation representing the vessel walls is generated using a direct tessellation of the boundary voxels. This surface is then smoothed and the quality of the resulting triangulation is improved. Finally, in the third stage, the triangulation is subdivided into so-called discrete surface patches for surface gridding, the desired element size distribution is de*ned and the *nite element grid generated