A Generalised Neural Network Model to Estimate Sex from Cranial Metric Traits: A Robust Training and Testing Approach

Abstract

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.

Mapping sexual dimorphism signal in the human cranium

“…When the variable cranial size is excluded from the analysis (shape space) the medial-inferior portion of the frontal bone resulted to be the most suitable cranial region for sex classification. Shape variations of this area agree with standard protocols commonly used in sex determination from cranial remains 12 , 31 , 46 . In females, the supraorbital torus is less developed, the glabellar region is more expanded, and the angle between the median and superior margin of the orbit is narrower than in males.…”

“…The scoring system used to characterize sex from cranial morphology takes into account variations in the size and shape of the glabella, mastoidal process, occipital crest, and orbital margin. On average, female individuals are smaller in cranial size and overall more gracile compared to males; however, only a few studies have focused on analyzing variations in cranial size at the interpopulation level 12 , 18 . In addition, the inclusion of size as a variable in sex identification leads to issues related to the overlapping of the signal due to both sexual dimorphism and cranial allometry.…”

“…Biological anthropology has been drastically changed thanks to the introduction of new methods of investigation. Since the foundation of virtual anthropology 19 , we are witnessing an increase in new strategies to identify sex by using geometric morphometrics methods 13 , 14 , 20 – 23 , beside the application of machine/deep learning 12 , 24 – 26 , 3D modeling techniques 27 , 28 and linear measurements 29 , 30 .…”

Artificial intelligence-based forensic sex determination of East Asian cadavers from skull morphology

“…Several studies have used machine-learning algorithms, a field of artificial intelligence, to evaluate the sex from skull shape and achieved high accuracies in forensic sex determination. We found two approaches in the literature: directly evaluating skull shape 10 or using Computed Tomography (CT) 11 – 15 . We suggest that the method using CT is superior to a direct evaluation in terms of convenience.…”