Jang, Jisun; (2025) Developing a novel adult age estimation method by calculating Dirichlet Normal Energy values from 3D reconstructed auricular surfaces. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Jang_10213120_thesis_sigs_removed.pdf Download (6MB) | Preview

Abstract

Estimating the age of human remains is crucial in bioarchaeology and forensic anthropology, but is largely dependent on methods that visually assess the degree and type of degeneration present in various regions of the skeleton and match this to broad age phases whose descriptions have been collated from observations of known-age skeletons. These methods have drawn criticism for their subjective nature and reliance on the experience of the observer in allocating an individual to a particular age phase, with more recent work focusing on developing computational methods which can identify and quantify age-related changes in a more objective and reproducible manner. This thesis used curvature analysis (Dirichlet Normal Energy (DNE)) to identify age-related patterns of curvature in the auricular surface of the ilium and presents a number of age prediction models which have been developed and tested using mathematical variables extracted from 3D reconstructed surfaces of known-age individuals from Europe (n=890), Southeast Asia (n=335), and post-medieval Britain (N=185). Three models (PCQDA, PCLR, and PCR) predict age with 76-93% accuracy for European individuals, 35-66% accuracy for Southeast Asian individuals, and 36-70% accuracy for post-medieval individuals. The results demonstrate that the auricular surface becomes more curved with increasing adult age in all three skeletal samples, although there are inter-population differences in the rate at which this occurs. The methodological procedure used to create the 3D models and prepare them for analysis has excellent rates of reproducibility (p(c) <0.82), suggesting that age predictions made with these models are objective and reliable. To address the problem of statistical bias relating to population affinity, two further models (A_PCR and M_PCR) were created using the Southeast Asian skeletal sample and a multi-population sample consisting of Southeast Asian and European individuals. Both models increased the accuracy of age estimations for Southeast Asians (69% and 67%, respectively), and post-medieval British individuals (41% and 42%, respectively). While the results display some degree of population affinity, the age prediction models can be applied to modern skeletal samples with high levels of accuracy, and may be applied to a wider variety of skeletal samples with an appropriate level of caution that acknowledges the population-specific biases. The thesis contributes to skeletal age estimation by introducing a computer-based method for detecting subtle auricular surface changes, eliminating subjectivity through DNE curvature analysis, and developing population-specific models to address statistical biases.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item