Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Mechanical and Materials Engineering


Dr. Thomas R. Jenkyn


Biomechanical investigations involving the characterization of biomaterials or improvement of implant design often employ finite element (FE) analysis. However, the contemporary method of developing a FE mesh from computed tomography scans involves much manual intervention and can be a tedious process. Researchers will often focus their efforts on creating a single highly validated FE model at the expense of incorporating variability of anatomical geometry and material properties, thus limiting the applicability of their findings. The goal of this thesis was to address this issue through the use of a statistical shape model (SSM). A SSM is a probabilistic description of the variation in the shape of a given class of object. (Additional scalar data, such as an elastic constant, can also be incorporated into the model.) By discretizing a sample (i.e. training set) of unique objects of the same class using a set of corresponding nodes, the main modes of shape variation within that shape class are discovered via principal component analysis. By combining the principal components using different linear combinations, new shape instances are created, each with its own unique geometry while retaining the characteristics of its shape class. In this thesis, FE models of the human craniofacial skeleton (CFS) were first validated to establish their viability. A mesh morphing procedure was then developed to map one mesh onto the geometry of 22 other CFS models forming a training set for a SSM of the CFS. After verifying that FE results derived from morphed meshes were no different from those obtained using meshes created with contemporary methods, a SSM of the human CFS was created, and 1000 CFS FE meshes produced. It was found that these meshes accurately described the geometric variation in human population, and were used in a Monte Carlo analysis of facial fracture, finding past studies attempting to characterize the fracture probability of the zygomatic bone are overly conservative.

morphing code.zip (14154 kB)