Mapping Lemur Teeth

A research group led by Prof. Ingrid Daubechies, and recent Duke PhD. graduate, Dr. Tingran Gao, in collaboration with the Duke Department of Evolutionary Anthropology, combines computer graphics, geometry processing, and machine learning, to study the evolutionary history of primates based on quantified variations of the shapes of anatomical surfaces like teeth and bones. By minimizing an energy functional between two triangular meshes, they defined a distance for two digitized anatomical surfaces, which also comes with a bijective smooth map between the pair. A machine learning framework is then developed, taking advantage of the pairwise distance and the bijective map, for clustering surfaces according to species groups, as well as partitioning all surfaces in the collection in a consistent manner. The research helps evolutionary anthropologists to reduce the manual work of the identification of cooresponding surface regions for large collections of fossil bones and teeth, and uncovered intriguing clustering patterns that reflect both the species and dietary groups.