Improving Chronological Age Estimation in Children Using the Demirjian Method Enhanced with Transformer and...

Area of Science:

• Forensic Anthropology
• Pediatric Dentistry
• Artificial Intelligence in Medicine

Background:

• Accurate chronological age estimation in children is crucial for legal and clinical contexts.
• Traditional methods for dental age assessment can be subjective and time-consuming.
• Deep learning offers potential for objective and efficient age estimation from radiographic data.

Purpose of the Study:

• To develop and validate a two-phase automated methodology for chronological age estimation in children.
• To leverage deep learning for feature extraction from panoramic dental radiographs.
• To compare the performance of various machine learning models for age prediction.

Main Methods:

• A dataset of 626 panoramic radiographs from children aged 6.0-13.8 years was used.
• Deep learning models (ResNet-18, EfficientNetV2-M, Swin V2 Base) extracted features from seven mandibular teeth.
• Extracted features were used to train nine machine learning regression models, with ExtraTrees showing the best performance.

Main Results:

• The ExtraTrees model achieved Root Mean Square Error (RMSE) of 6.98 months for females and 6.55 months for males.
• Mean Absolute Error (MAE) was 5.18 months for females and 5.01 months for males.
• SHAP analysis identified the first premolar (P1) and second molar (M2) as key features for age prediction.

Conclusions:

• The proposed automated pipeline significantly enhances the accuracy of dental age estimation.
• This method reduces observer variability compared to traditional techniques.
• The developed tool provides a reliable and efficient solution for clinical and forensic dental age assessment.