The study of sexual dimorphism in human crania has important applications in the fields of human evolution and human osteology. Current, the identification of sex from cranial morphology relies on manual visual inspection of identifiable anatomical features, which can lead to bias due to user’s expertise. We developed a landmark-based approach to automatically map the sexual dimorphism signal on the human cranium. We used a sex-known sample of 228 individuals from different geographical locations to identify which cranial regions are most sexually dimorphic taking into account shape, form and size. Our results, which align with standard protocols, show that glabellar and supraciliary regions, the mastoid process and the nasal region are the most sexually dimorphic traits (with an accuracy of 73%). The accuracy increased to 77% if they were considered together. Surprisingly the occipital external protuberance resulted to be not sexually dimorphic but mainly related to variations in size. Our approach here applied could be expanded to map other variable signals on skeletal morphology.

The morphology of the human cranium allows for reconstructing important information about the identity of an individual, such as age, ancestry, sex, and health status. The estimation of sex from morphology is a key component of the work of physical anthropologists, and in the last decade, the field has witnessed an increase in the use of novel algorithm-based methodologies to tackle the aforementioned task. Nevertheless, several limitations (e.g., small training/testing sample size, training-test data relatedness, limited population inclusiveness, overfitting) have hampered the application of such methods as a standardised procedure in the field. Here, we propose a population-inclusive protocol for estimating sex from a small set of cranial metric traits (10 measurements) based on a neural network architecture trained to maximise the probability of sex attribution and prevent overfitting. The cross-validation returned an accuracy of 86.7% ± 0.02% and log loss of 0.34 ± 0.03. The protocol developed was tested on data unrelated to that of the training and validation phase and returned an estimated accuracy of 84.3% and log loss of 0.348. The model and the related code to use it are made publicly available.

The observation and the quantification of asymmetry in biological structures are deeply investigated in geometric morphometrics. Patterns of asymmetry were explored in both living and fossil species. In living organisms, levels of directional and fluctuating asymmetry are informative about developmental processes and health status of the individuals. Paleontologists are primarily interested in asymmetric features introduced by the taphonomic process, as they may significantly alter the original shape of the biological remains, hampering the interpretation of morphological features which may have profound evolutionary significance. Here, we provide a new R tool that produces the numerical quantification of fluctuating and directional asymmetry and charts asymmetry directly on the specimens under study, allowing the visual inspection of the asymmetry pattern. We tested this show.asymmetry algorithm, written in the R language, on fossil and living cranial remains of the genus Homo. show.asymmetry proved successful in discriminating levels of asymmetry among sexes in Homo sapiens, to tell apart fossil from living Homo skulls, to map effectively taphonomic distortion directly on the fossil skulls, and to provide evidence that digital restoration obliterates natural asymmetry to unnaturally low levels.

Skeletal remains analyzed by anthropologists, paleontologists and forensic scientists are usually found fragmented or incomplete. Accurate estimations of the original morphologies are a challenge for which several digital reconstruction methods have been proposed. In this study, the accuracy of reconstructing bones based on multiple linear regression (RM) was tested. A total of 150 digital models from complete zygomatics from recent past populations (European and African American) were studied using high-density geometric morphometrics. Some landmarks (i.e., 2, 3 and 6) were coded as missing to simulate incomplete zygomatics and the missing landmarks were estimated with RM. In the zygomatics, this simulated damage affects a few square centimeters or less. Finally, the predicted and original shape data were compared. The results indicate that the predicted landmark coordinates were significantly different from the original ones, although this difference was less than the difference between the original zygomatic and the mean zygomatic in the sample. The performance of the method was affected by the location and the number of missing landmarks, with decreasing accuracy with increasing damaged area. We conclude that RM can accurately estimate the original appearance of the zygomatics when the damage is small.