A Methodological Study On The Taxonomic Identifications Of Hominid Teeth

Teeth are important information carriers for exploring the origin,evolution and spread of hominid.Nevertheless,for specific taxa(genera or species),their teeth could show morphological similarities or overlapping on the quantitative parameters,making dental taxonomic discrimination a challenging task in the paleoanthropology field.Therefore,it is necessary to develop more reliable indexes or methods to identify hominid teeth.For this purpose,efforts can be started from two perspectives:1.improvement of traditional methods;2.introducing relevant methods from other disciplines.Three-dimensional relative enamel thickness(3DRET)is an important traditional index for taxonomic identification of hominid teeth.Micro-CT is now a routine way to measure 3DRET.Here,3DRET was shown to have two flaws:1.this index is sensitive to the voxel size of CT image;2.there exists a linear relationship between 3DRET and three-dimensional average enamel thickness(3DAET).The former flaw prevents the effective comparison of different studies(for a specific genera or species)where various voxel sizes are likely to be adopted to reconstruct the virtual teeth models,while the latter can lead to information redundancy.For these reasons,we propose an alternative to 3DRET.The new index is defined as the ratio of enamel-thickness to dentine-thickness(3DRED).To compare the sensitivity(to voxel size)of 3DRET,3DRED and3DAET,a fossil orangutan molar was scanned 14 times with different resolutions ranging from 10 to 50?m.Enamel thickness analysis was carried out for each resultant digital model.In addition,enamel thickness measurements of 179 mandibular permanent molars(eight genera,including Neanderthals,Modern humans,Paranthropus robustus,Pan troglodytes,Gorilla gorilla,Hylobates sp.,Symphalangus syndactylus,and Pongo sp.)were analyzed,followed by investigating the relationship between 3DRET and 3DAET and between 3DRED and 3DAET.The results show that:1.regarding sensitivity,3DRED is more robust than 3DRET;2.3DRET is correlated with 3DAET by linear curve with regression coefficients approximating or larger than0.8 in most cases,while 3DRED shows less correlation with 3DAET(regression coefficients are less than 0.5 in most cases);3.For the mandibular first molars(M₁)and second molars(M₂),there are clear separations between different taxa in the bivariate plot of 3DRED against 3DAET,indicative of the taxonomic value of 3DRED.Compared to 3DRET,3DRED is more robust and reliable in taxonomic study.Convolutional neural network(CNN)is a state-of-art deep learning(DL)method with superior performance in image classification.Here,for the first time,we introduce CNN into the paleoanthropology field and a CNN-based workflow is proposed to discriminate hominid teeth.The first step is converting the enamel-dentine junction(EDJ)into EDJ card,i.e.,a two-dimensional(2D)image conversion of the three-dimensional EDJ surface.Specifically,the EDJ card is obtained via projecting the EDJ surface to the cervical plane and expressing the height pattern in a form of colormap.The second step is training CNN learner with the labeled EDJ cards.A sample consisting of 53 fossil Pongo(original data)and 53 Homo(modern humans and Neanderthals,most of them are from open source CT database)was adopted to generate EDJ cards,which were then separated into training set(n=84)and validation set(n=22).To assess the feasibility of this workflow,a Pongo-Homo classifier was trained from the aforementioned EDJ card set,and then the classifier was used to predict the taxonomic affinities of six samples(test set)from von Koenigswald's Chinese Apothecary collection.Results show that EDJ cards in validation set are classified accurately by CNN.More importantly,taxonomic predictions for six specimens in test set match well with the diagnosis results deduced from multiple lines of evidence.The above results are achieved from small sample training,implying the great potential of CNN method.The postcanine of Gigantopithecus blacki from Chuifeng and Mohui cave were adopted to study the taxonomy,phylogeny and dietary of G.blacki.This case study is not only used to assess the performance of the two methods described above,but also aimed at sharping the understanding of the dietary ecology of this extinct great ape.For that,three studies were carried out,respectively.First,six slightly worn mandibular molars were selected for enamel thickness analysis,and the results show that the3DAET-3DRED bivariate plot is effective in distinguishing G.blacki from other taxa.Then,the EDJ card of a maxillary first/second molar was generated,and this card was taken as the input to determine the phylogenetic status of G.blacki using the trained Pongo/Homo classifier.DL results suggest that the EDJ morphology of G.blacki is more Pongo-like.Finally,sixteen permanent mandibular teeth were employed to assess bite force(BF).The results show that the average molar BF is 1512N,which is larger than taxonomically broad primate taxa.Besides,G.blacki shows a distal-ward reduction in BF,with the average BF of premolar(2217N)being approximately 1.5times that of molar.In a new way,the great BF strengthened previous consensus that G.blacki is capable of consuming food that hard to crush/abrade.The specialization,significant BF difference between premolar and molar,is most likely to be an ecologic adaptation to the intake of bulky hard food items.