Machine learning in sex estimation using CBCT morphometric measurements of canines

Area of Science:

• Forensic Anthropology
• Dental Forensics
• Machine Learning in Medicine

Background:

• Accurate sex estimation is crucial in forensic investigations.
• Dental structures, particularly maxillary canines, offer potential for biometric analysis.
• Cone Beam Computed Tomography (CBCT) provides detailed 3D imaging for dental measurements.

Purpose of the Study:

• To evaluate maxillary canine measurements obtained via CBCT for sex estimation.
• To develop and validate machine learning models for automated sex determination using dental metrics.

Main Methods:

• CBCT scans of 610 patients were analyzed for maxillary canine dimensions (total length, enamel thickness, mesiodistal width).
• Supervised machine learning algorithms (e.g., LightGBM, Logistic Regression, Random Forest) were trained and validated using 10-fold cross-validation.
• Performance was assessed using metrics including Area Under the Curve (AUC), accuracy, precision, and F1 Score.

Main Results:

• Maxillary canine total tooth length was the most significant variable for sex estimation.
• LightGBM and Logistic Regression models achieved the highest AUC values (0.77 and 0.75, respectively) on test data.
• Both top-performing models demonstrated high precision, indicating reliable predictions.

Conclusions:

• Maxillary canine measurements, when analyzed with machine learning, provide a viable method for sex estimation.
• This approach offers a low-cost, single-anatomical-structure-based solution for forensic identification.