Parameterised reference models for hominoid skulls

For the field of anthropology, the discrimination of taxa and the qualitative and quantitative description of ontogenetic or evolutionary change of form and shape is a fundamental area of research. The conventional metrical approach is known to be inherently deficient and the results are often ambiguous and contentious. Driven by the increasing availability of modern 3D-digitising technology and the related enormous amount of information available, methods based on 3D-coordinates of landmarks or semi-landmarks, on information of curvature, or on the topology of surfaces make quantifying morphology possible. With CT-scans of modern and fossil skulls, intracranial and hidden structures can now be studied with an accuracy comparable to those structures on the visible surface and provide new information for phylogenetic classification. On the specimen`s way from biosphere to lithosphere, destruction and distortion frequently take place. Fossils are usually found broken into many pieces and researchers need reconstruction. In most cases, this problem is left to the specialist with extensive experience and a well-trained `morphological eye`. A more reproducible methodology which would also offer the opportunity for repeated attempts is reassembling of fragments in a computer environment, based on quantitative morphology data and on statistical knowledge about form and shape of the species to be reconstructed. In our project, it is intended to create parameterised reference models of skulls of different species for two purposes: 1. To describe the deviations of a specimen from this average form by its change of parameters (morphological comparison), and 2. to attach fragments electronically according to a best fit criterion to this parameterised reference model of an organism (fragment reassembly). Landmark descriptors (traditional, Procrustes-, and Thin Plate Spline-measures), differential geometry descriptors (extremal points, ridge curves, semi-landmarks), and surface descriptors (coefficients of surface approximation) will be used to characterise a set of skulls. From these defining information structures, `hyperparameters` will be computed which include features of local and global variation.

Fossil bones millions of years old are only rarely preserved completely and in good condition. Also, the changes of skull form over the course of evolution or individual growth are quite difficult to reduce to numbers using the traditional standard methods. Under the framework of this three-year FWF project, our team of scientists in the Department of Anthropology of the University of Vienna has developed many new methods for computerbased preparation, reconstruction, and geometrical assessment of skulls, methods with potential applications in medicine and industry as well. Skull bones that deformed like rubber under the pressure of sediments can be restored to their original form. Their cavities can be "virtually" freed of stony deposits [encrustations], and missing pieces can be put back by reflecting or warping the parts that are there. The most modern techniques of 3D digitising and computed tomography and our new algorithms and software make all this much more exact than was previously possible. "Biological forms are very complex objects, which can`t be described by simple functions as if they were car bodies," explains Univ. Prof. Gerhard W. Weber, leader of the working group "Virtual Anthropology" in the department. To characterize the structures by numbers and pixels, the scientists developed complicated mathematical methods based on a large number of homologous reference points on the object surface. With the help of these reference points (landmarks and semilandmarks) it is possible, among other things, to consider shape separately from size. One of the main advantages is doubtless the reproducibility of evidence gathered this way. Up to now the comparison and reconstruction of bony discoveries was reserved for a few specialists and their educated but consequently subjective "morphological eye." Now what counts are quantitative data and the statistical understanding of shape and form. The net spatial information about an object can be processed all at the same time with an intensity not hitherto possible, and form comparisons can be visualized in space as distorted coordinate systems. Further applications to medicine are already in progress with colleagues at the University of Innsbruck. They are loading 3D images of lungs into computers, one on top of another, and investigating how the contents of the lung depend on the details of artificial respiration.