FROM MICROFOSSILS TO MOLARS: TESTING THE APPLICATION OF THE SOFTWARE PACKAGE AUTOMORPH TO FOSSIL MAMMAL TEETH

Teeth represent a significant portion of the fossil mammal record and are often useful for assigning taxonomic determinations of a specimen. However, without a high degree of familiarity, teeth are often difficult to identify due to small differences in tooth morphology among mammal species. AutoMorph, a software package developed in 2017 by Hsiang and colleagues, may provide a semi-automated method to determine the taxonomic identity of fossil mammal teeth. AutoMorph is an open-source, high-throughput software package that can rapidly extract 2D and 3D morphological data from both standard and focal-stacked images. This software was developed to automate the taxonomic identification of foraminifera from deep-sea sediments, and has also successfully been used to identify marine invertebrates and ichthyoliths. To date, no published works have applied AutoMorph to the fossil mammal tooth record. Here, we present preliminary tests exploring the ability of AutoMorph to extract morphological data from small mammal teeth and the significance of inter- and intra-user variation in imaged tooth specimens (repeatability). Limiting our test specimens to Hyopsodus (Condylarthra, Hyopsodontidae) from the University of Wyoming Fossil Vertebrate Collections, we imaged the occlusal surfaces of lower molars. AutoMorph processed these images and returned measurements for nine 2D variables, including axis lengths, aspect ratio, area, and perimeter. Results show that, with minimal post-processing, AutoMorph is able to quickly and efficiently extract 2D data from images of fossil mammal teeth. Comparisons of AutoMorph’s extracted axis lengths to caliper and digital microscope measurements indicate that AutoMorph provides statistically indistinguishable, high-fidelity measurements. Calculations of mean error for intra- and inter-user variation for AutoMorph measurements show <1% error for all returned variables, with the exception of the perimeter variable (error=1.48%). These results indicate that small differences in the initial image capture should not significantly impact AutoMorph’s analyses. Our findings provide the first step towards applying AutoMorph to studies of fossil mammal teeth, and this software package will likely be useful for future identification and comparative studies.