Age estimation plays a significant role in forensic anthropology and bioarchaeology. However, widely-used traditional methods involving macroscopic observation suffer from subjectivity and statistical bias. The present research aims to minimize both issues by applying computational and mathematical approaches. A laser scanner was used to reconstruct 890 auricular surfaces of adult individuals from three known-age European skeletal collections. Dirichlet Normal Energy (DNE) was applied to assess the curvature of the auricular surface and its relationship with known age-at-death. Ten variables had high correlations, including total DNE per Total polygon faces, Mean value of DNE on apex, proportion of polygon faces with DNE of less than 0.0001 and proportion of polygon faces with DNE of over 0.6. The variables were used to develop age prediction models which are freely available in a novel R package, JSDNE. The package predicts age mathematically, objectively, and user-independently. It includes three functions: principal component quadratic discriminant analysis (PCQDA), principal component regression analysis (PCR), and principal component logistic regression analysis (PCLR), which produce age estimates with 91%, 76%, and 92.9% levels of accuracy, respectively. JSDNE package (https://cran.r-project.org/package=JSDNE) can be downloaded automatically using install.packages("JSDNE"). The detailed code and the raw data of this study are openly available at https://github.com/jisunjang19/cran-JSDNE, doi: 10.5281/zenodo.12708779 or ‘JSDNE’ package.
Read full abstract