The utility of elliptical Fourier analysis for estimating ancestry and sex from lateral skull photographs

Abstract

Current quantitative methods for estimating ancestry and sex from skulls typically require substantial manual data collection and specialized recording equipment, which can limit analysis to the laboratory. This limitation could be addressed by establishing a faster, more user-friendly, and automatic data protocol as investigated in the current study using elliptical Fourier analysis (EFA). Ancestry and sex were estimated using outlines acquired from standardized photographs of the skull in norma lateralis (left side). In this investigation, training samples comprised anatomical specimens from five collections: the Hamann-Todd Human Osteological Collection, WM Bass Donated Skeletal Collection, Robert J. Terry Anatomical Skeletal Collection, Khon Kaen Osteological Collection, and Chiba Bone Collection. Groups were defined as Black American female (n=87), Black American male (n=109), Japanese male (n=59), Thai female (n=39), Thai male (n=47), White American female (n=97), and White American male (n=134). EFA was conducted on partial Procrustes-aligned skull outline coordinates, before extracting principal components and using linear discriminant analysis for group assignment. Classification accuracy was determined using the 5-fold cross-validation protocol. Ancestry and sex were classified correctly 73% of the time when all seven reference samples were used. When only Black and White Americans were retained in the reference sample with sex pooled, they were correctly classified 94% of the time. Accuracy of out-of-group ancestry and sex estimation was evaluated using nine White American males from the Defense POW/MIA Accounting Agency Laboratory. A seven-way comparison with all reference samples for estimating both ancestry and sex achieved 89% (8/9) correct classifications, with one misclassification as White American female. These out-of-group results, along with initial training group accuracies, indicate that lateral skull outlines can be used to successfully estimate ancestry and sex with similar accuracy to other methods, and set the basis for future cross-validation testing. Further, the reliance on a single easy-to-take photograph and user-friendly open-source R script facilitates easy application and field use. The protocol is freely available from CRANIOFACIALidentification.com as the SkullProfiler script.