Developing the standards used in facial approximation

: a linear measurement & geometric morphometric investigation of the eye in a South Korean Population

Abstract

This study aimed to refine and develop methodologies for predicting the soft tissue appearance of the eye based on skeletal structures, with a focus on forensic facial approximation within the South Korean population. Utilizing 66 Cone Beam Computed Tomography (CBCT) scans- comprising 33 male and 33 female subjects aged 20 to 45- systematic measurements of soft and hard tissues were conducted employing the DICOM viewer and the 3D Slicer application. A semi-automated Python script facilitated the extraction of anatomical landmarks and linear measurements, thereby ensuring efficient data collection. The analytical framework encompassed descriptive analysis, repeatability assessments, T-tests, normality evaluations, Pearson correlation, linear regression analysis, and validation analysis. The Geometric Morphometric analysis within MorphoJ utilized Procrustes superimposition, Principal Component Analysis, and Procrustes ANOVA to investigate relationships between hard and soft tissues. The findings confirmed the validity of pupil-to-orbital margin measurements, indicating inconsistencies in the positioning of the canthi. Linear regression analysis yielded 28 regression formulas employed to predict the soft tissue characteristics of the orbital region based on the underlying hard tissue. The geometric morphometric analysis highlighted that sex significantly influences orbital size and shape, whereas age exhibits a lesser effect on soft tissue variation. Moreover, genetic factors affecting epicanthic fold types were observed. The study posits that expanding datasets and refining predictive models may enhance the accuracy and reliability of these techniques in future investigations. Facial approximation is widely used in forensic contexts to generate investigative leads when conventional identification methods fail, making these refinements valuable for real-world investigations. Both linear and Euclidean distances were extracted for analysis. Measurements were taken on segmented, 3D-rendered cranial and facial models created in 3D Slicer.

Date of Award	2025
Original language	English
Awarding Institution	University of Dundee
Sponsors	National Forensic Service
Supervisor	Tobias Houlton (Supervisor) & Fordyce Davidson (Supervisor)