Automated dental age estimation model for panoramic radiographs

Developmental stages of the dentition and age-group classification are crucial in forensic odontology, oral radiology, pediatric dentistry, and orthodontic dentistry. However, the manual interpretation of dental panoramic (OPG) radiographs for assessing the developmental stages and further age-group classification is time-consuming and prone to variability. This current model aims to evaluate the performance of a computational neural network (CNN) based model for automated detection of Nolla's developmental stages of permanent teeth and subsequent dental age group classification using digital OPGs from individuals aged 3 - 30 years.

Nolla's method was selected for its broad applicability to all the permanent teeth, including the developing third molar, and its comprehensive staging system. A total of 4073 radiographic images for Nolla's developmental stages (0 to 10) and age-group classification were utilized for training and validation of an annotated dataset, with an independent dataset of 1450 anonymized radiographs with varying quality and 644 repository images for testing of the YOLOv8-based models.

For the developmental stage classification, the model achieved a precision of 0.912, recall rate of 0.927, F1 score of 0.919, and a mean average precision (mAP) of 0.963. For age-group classification, the model demonstrated a precision of 0.83, recall of 0.79, F1 score of 0.81, and mAP of 0.88. The overall performance was highest in the mid-age range, and the lower metrics were observed in younger age groups less than 6 years of age.

The current model effectively identifies and classifies the developmental stages of permanent teeth and estimates age from panoramic radiographs with high accuracy, especially in well-represented age groups. It emphasizes the feasibility of applying real-time object detection frameworks for age assessment. While not intended as a validated forensic age-group classification tool, the model provides a proof-of-concept for integrating real-time object detection approaches into dental developmental assessment workflows and establishes a foundation for future population-specific, sex stratified, and forensically validated models.