Anthropology, University of Freiburg, Germany
Saturday All day, Clinch Concourse
In Geometric Morphometrics, most methods are applied on sparse sets of homologous landmarks, manually placed on the structures to be analysed. The drawbacks of manual landmark placement are evident: though modern digital imaging provides very accurate surface representations of biological structures, there are few landmarks that are well-defined and reliable in terms of homology and observer error. As a consequence, the rich information contained in the digital informations is poorly exploited. Many approaches to overcome these shortcomings (e.g. sliding semi-landmarks), still rely more or less on manually placed landmarks, thus being subject to observer error. We are presenting a method using smoothed displacement fields to register an atlas surface onto all other surfaces in our sample. Hereby, manual interaction is only required to place reference points establishing an initial, coarse spatial correspondence. To test the performance of our proposed method, we choose a rather complex geometric structure: the human zygomatic bone. In our study, we apply our registration method to 200 surface meshes extracted from CT-scans in order to evaluate population- and gender-specific surface shapes. 8 landmarks provide information for the surface meshes to be coarsely aligned, followed by an enhanced iterative closest point rigid body registration. The aligned surfaces now are iteratively matched onto each other using a locally smoothed displacement vector field. Finally, all surfaces are represented by meshes, consisting of corresponding vertices. Vertices within the region of interest can be processed, using the well established methods of Geometric Morphometrics.