Machine Learning for Estimating Chronological Age in Children Using the London Atlas and Dental Radiographs

This study aims to evaluate convolutional neural network models for automated estimation of chronological age in children using dental radiographs. A dataset of 28 pediatric dental radiographs were collected from individuals with known chronological ages. Using the London Atlas, dental development stages were annotated. Before training the CNN models, preprocessing and segmentation techniques were applied. For supervised learning models were used: MobileNetV2, DenseNet121, EfficientNet B0 and ResNet34. Performance was evaluated using parameters, such as accuracy, precision, sensitivity and specificity. Among the models, the MobileNetV2 architecture achieved the best performance, with high correlation between predicted and actual age with 95% accuracy. The integration of machine learning with dental radiographic analysis provides a promising tool for accurate, age estimation. This approach may enhance diagnostic workflows in pediatric dentistry and forensic identification practices.