Automated dental age estimation model for panoramic radiographs.

Area of Science:

• Forensic Odontology
• Radiology
• Pediatric Dentistry
• Orthodontic Dentistry

Background:

• Accurate dental development assessment is vital for various dental specialties.
• Manual interpretation of dental panoramic radiographs is time-consuming and variable.
• Automated methods are needed to improve efficiency and consistency.

Purpose of the Study:

• To evaluate a convolutional neural network (CNN) model for automated detection of Nolla's developmental stages.
• To assess the model's capability for subsequent dental age group classification using digital panoramic radiographs.
• To analyze the performance of the model in individuals aged 3-30 years.

Main Methods:

• Utilized Nolla's comprehensive staging system for permanent teeth development.
• Trained and validated a YOLOv8-based CNN model on 4073 annotated panoramic radiographs.
• Tested the model on an independent dataset of 1450 anonymized and 644 repository images.

Main Results:

• The model achieved high precision (0.912), recall (0.927), and F1 score (0.919) for developmental stage classification.
• Age group classification yielded precision (0.83), recall (0.79), and F1 score (0.81).
• Performance was optimal in mid-age ranges, with lower accuracy in children under 6 years.

Conclusions:

• The CNN model accurately identifies permanent teeth developmental stages and estimates age from panoramic radiographs.
• Demonstrates the feasibility of real-time object detection for dental age assessment.
• Provides a proof-of-concept for integrating automated tools into dental developmental assessment, paving the way for validated forensic models.