Sex Estimation Through Orbital Measurements: A Machine Learning Approach for Forensic Science.

BACKGROUND

Sex estimation has been extensively investigated due to its importance for forensic science. Several anatomical structures of the human body have been used for this process. The human skull has important landmarks that can serve as reliable sex estimation predictors.

METHODS

In this study, orbital measurements from 92 dried skulls, comprising 35 males and 57 females, were utilized to develop a machine-learning-based classifier for sex estimation with potential applications in forensic science. The parameters evaluated included optic foramen height (OFH), optic foramen width (OFW), optic canal height (OCH), optic canal width (OCW), intraorbital distance (IOD), extraorbital distance (EOD), orbit height (OH), and orbit width (OW).

RESULTS

A Random Forest classifier was employed to analyze the data, achieving an overall test accuracy of 0.68. The model demonstrated a precision of 0.65, indicating a moderate level of false positives. The recall was 0.70, reflecting that 70% of the positive cases were correctly identified. The F1 score was calculated at 0.675, suggesting a balanced performance between precision and recall. The area under the curve (ROC AUC) score was also 0.72, indicating that the model can distinguish between classes. The most important features in the best subset were OW (0.2429), IOD (0.2059), EOD (0.1927), OFH (0.1798), and OFW (0.1787), highlighting their significant contributions to the model's predictions.

CONCLUSIONS

These findings suggest that orbital measurements could potentially serve as reliable predictors for automated sex estimation, contributing to advancements in forensic identification techniques.