Evaluating the Effectiveness of Coxal Bone Measurements for Sex Estimation via Machine Learning

Area of Science:

• Forensic Anthropology
• Human Anatomy
• Biometrics

Background:

• The human pelvis exhibits significant sexual dimorphism, largely due to its role in childbirth.
• Sexually dimorphic traits are particularly prominent in the coxal bones, crucial components of the pelvic girdle and birth canal.

Purpose of the Study:

• To quantify sex differences in coxal bone dimensions.
• To develop and evaluate machine learning models for sex estimation using coxal bone morphometrics.

Main Methods:

• Utilized computed tomography (CT) scans of 276 adult Bulgarian individuals.
• Generated 3D pelvic models and collected 34 landmark coordinates from coxal bones.
• Calculated various coxal bone measurements and analyzed differences based on sex, age, and laterality.
• Trained machine learning models (Support Vector Machines, logistic regression) for sex classification.

Main Results:

• Demonstrated significant sexual dimorphism in coxal bone dimensions.
• Identified minor bilateral and age-related variations in pelvic morphology.
• Achieved high accuracy (95-100%) in sex estimation using the developed machine learning models.

Conclusions:

• Coxal bone morphology displays pronounced sexual dimorphism, making it a reliable indicator for sex estimation.
• Machine learning approaches effectively utilize coxal bone measurements for accurate sex determination in forensic and anthropological contexts.