抄録
Identifying individual tooth (tooth class and type within it) is a fundamental skill for anthropologists. It is always required for basic description in the fieldwork. Besides, it is vital to identify dental specimens in the studies to infer phylogenetic relationships and species classifica-tion. Likewise, the identification of teeth is highly relevant in the primary education of dentistry. However, the characteristics described in standard textbooks do not necessarily apply to all teeth due to broad inter-individual variation of morphology. While the qualitative description of dental traits is useful, it better be constructed on the ground of quantitative analysis. Here, we introduced a method that combines the technique of geometric morphometrics and machine learning. Specifi-cally, we used the method of morphometric mapping to quantify and visualize three-dimensional tooth crown morphology of human upper molars and to extract multiple morphological parameters which can then be submitted to machine learning. Results show that the classification accuracy is maximized when using the x component of vertex normal toward mesio-distal direction with a small filter size for noise reduction. The mesio-distal gradient of tooth crown morphology is highly relevant for molar type identification with algorithmic processing, which is underpinned by the morphogenetic process of tooth formation.
https://ror.org/04r8tsy16 
