Abstract:
Estimating sex from unknown human skeletal remains is an important component in forensic anthropology. Currently, both morphological and morphometric methods are used for sex estimation. These methods employ landmarks to make morphological comparisons between and within groups. Manual landmarking has been regarded as time-consuming and subjective. To decrease observer subjectivity and reduce measurement errors, an automated three-dimensional (3D) method was developed. This study aimed to validate the utilisation of the automatic placement of anatomical and sliding landmarks on 3D pelvis models for shape analysis using Computed Tomography (CT) scans.
Computed Tomography scans of adult South Africans were obtained from Steve Biko Academic Hospital, Pretoria, South Africa. In this study, automatic landmarking was validated on 130 3D reconstructions of the adult human pelvis. Eighteen anatomical and 260 sliding landmarks were registered on 130 3D models of the same individuals manually and automatically using the MeVisLab © v 2.7.1 software. Landmark datasets were acquired using both landmarking methods and compared using reproducibility testing and geometric morphometric (GMM) analysis. Reproducibility testing of both landmark datasets demonstrated minimal dispersion errors (<2 mm), indicating the reliability and repeatability of both landmarking methods. Variance analysis showed that pelvis shape sex-related variation was statistically significant (p <0.05) using both methods. In addition, cross-validated discriminant function analysis (DFA) yielded accuracies between 82.98 – 97.73% and 65.91 – 93.18% using automatic and manual placement of landmarks, respectively.
In forensics using 3D automatic approaches, and advanced statistical analysis might allow forensic anthropologists to estimate sex in a more accurate and repeatable way.
