Level set methods for computational prototyping with application to hemodynamic modeling

Geometry is a primary determinant of hemodynamic conditions, including complex recirculating flow patterns which are hypothesized to be responsible for the localization of vascular disease and failure of surgical interventions. Computational flow simulations using subject-specific geometric models can play an important role in vascular disease research and operative planning, but accurate models are essential. The goal of this research is to develop efficient methods for constructing accurate three-dimensional subject-specific models of complex vascular structures suitable for numerical analysis based on volumetric medical images.; Because geometric artifacts can lead to non-physical features in a numerical solution, it is desirable for vascular models to be smooth. In order to provide for such smoothness, a geometric image analysis approach based on the level set method is developed. This approach consists of convergent behavior and precise smoothness constraints. The parameterization of the method is considered, and two criteria for detecting boundary convergence are described. In addition, to reduce the memory requirements associated with these calculations, a data structure for compact storage of level set grids is also presented. This technique is applied in conjunction with a profile-based strategy for constructing analytic surfaces, and several key issues related to modeling tortuous vascular structures are discussed. The use of cross-sectional images normal to a vessel's axis is motivated and described, and a strategy for reduction of parametric twist in analytic surfaces is given.; A modular software architecture for image-based model construction is also developed. Characteristics of this architecture include an object-oriented a shared repository mechanism for inter-module communications and a high-level scripting language for front-end integration.; The image analysis and model construction methods of this work are validated using a new geometric segmentation quality metric and two types of test images. Then, these techniques are applied to the construction of several subject-specific computational models, and a finite element blood flow calculation is described.