Using Kvaal method and machine learning to improve adult dental age estimation with CBCT

Area of Science:

• Forensic Anthropology
• Radiology
• Machine Learning

Background:

• Accurate adult age estimation is crucial in forensic science.
• The Kvaal method is a conventional technique for dental age estimation.
• Cone-beam computed tomography (CBCT) offers detailed dental imaging.

Purpose of the Study:

• To enhance dental age estimation accuracy using CBCT and machine learning.
• To compare machine learning models with traditional regression methods.
• To evaluate the effectiveness of the Kvaal method with advanced computational techniques.

Main Methods:

• Analysis of CBCT scans from 400 Northern Chinese individuals (aged 21-70).
• Measurement of Kvaal-derived indices for selected teeth.
• Development and comparison of sex-specific linear regression, Random Forest (RF), and Extreme Gradient Boosting (XGBoost) models.

Main Results:

• Width-related dental indices showed stronger age correlations than length-related ones.
• Machine learning models (RF and XGBoost) significantly reduced Mean Absolute Error (MAE) by 15-25% compared to linear regression.
• Best MAE achieved was 6.72 years (RF, male maxillary second premolars) and 7.18 years (XGBoost, male mandibular lateral incisors).

Conclusions:

• CBCT-based dental age estimation integrated with machine learning offers modest accuracy improvements over the conventional Kvaal method.
• This combined approach better captures age-related dental changes for more reliable forensic age estimation.
• Further validation in diverse populations is recommended for broader forensic application.