Measuring the inter-species variability of endocast growth using shape regression and spatiotemporal registration

Large database of anatomical surfaces offer the way for an automatic and quantitative analysis of the morphology of population and species. In this study, we used fossil endocasts segmented from CT images of chimpanzees and bonobos, as the closest extant relatives of modern humans. Given the dental age of every observation, our purpose is to analyze not only the morphological variations between species at a given age, but also the difference in terms of the growth of the endocasts. For this study, we applied advanced morphometric tools introduced recently by the authors in Computational Anatomy. Contrary to standard morphometric techniques, we analyze the global shape of the endocast without relying on specific landmarks. Surfaces are modeled as currents, which enables surface comparison in absence of point correspondences between surfaces. Large deformations are used to capture anatomical variations. First, we performed a shape regression (shape vs. dental age), which infers a 'typical' continuous evolution from the endocasts of the same species. Then, given two growth scenarios, we used spatiotemporal registration to analyze the differences between both scenarios in terms of morphological changes (at a given age) and change of the growth speed (once the morphological changes have been discounted). Eventually, this study shows the potential of new methodological tools to give a visual description of the inter-species variability and to support it with quantitative measures. Possible applications are a better determination of the age of a new endocast and a better understanding of the main shape differences between species.