The 82nd Annual Meeting of the American Association of Physical Anthropologists (2013)

Automated approaches to geometric morphometrics


1Program In Applied and Computational Mathematics, Princeton University, 2Department of Evolutionary Anthropology, Duke University, 3NYCEP, New York Consortium in Evolutionary Primatology, 4Dept of Anthropology, CUNY Graduate Center, 5Mathematics Department, Duke University

Friday 45, Clinch Concourse Add to calendar

Three-dimensional geometric morphometric (3DGM) methods for placing landmarks on digitized bones have become increasingly sophisticated, including algorithms that optimize the distribution of semi-landmarks within user-defined patches. One aspect that all 3DGM methods share is that initial landmarks must be designated by the researcher.Thus researcher interpretations of homology and correspondence influence, and are required for, representations of shape.

We present new algorithms allowing fully automatic placement of correspondence points on samples of 3D digital models representing bones of different individuals/species which can then be input into standard 3DGM software and analyzed with dimension reduction techniques. We test this algorithm with a sample of 106 primate calcanei represented by 1,200 correspondence points per bone. For comparison, we generated a traditional 3DGM dataset on the same sample, placing 27 landmarks on each bone. Data were analyzed with morphologika2.5. Initial results show strong similarities between the shape spaces generated by the automatic and traditional methods. PC1 corresponded mainly to distal calcaneal elongation in both plots. PC2 corresponded visibly to position and size of the peroneal process. In both analyses PC1 and PC2 together account for ~50% of the total sample variance. In plots of PC1 with PC3, both datasets showed almost perfect separation of major primate clades. These results indicate that automatic quantifications can lead to shape space representations that are at least as meaningful as those based on observer landmarks, thus providing the potential to save time in data collection while also increasing completeness of morphological quantification and eliminating observer error.

This research was supported by an NSF grant to DMB (NSF BCS 1125507)

comments powered by Disqus