AI-assisted age estimation in children based on a combination of bone and tooth maturity

Background: International protocols for age estimation in subadults recommend combining different evidence according to tooth and bone maturity by radiographic examination to improve the final assessment. Scant literature could be found that observe, compare, and combine dentition and wrist bones maturation in the same sample of minors. Aim: This research aims at developing and validating an Artificial Intelligence (AI)-assisted method combining the skeletal and dental methods for age estimation in children and adolescents. Material and methods: The sample consisted of orthopantomography and wrist radiographs of 453 Italian subadults (227 males and 226 females) taken for clinical reasons. The age of the sample group is between 6 and 20 years old. The dental age was estimated by applying Demirjian 7-teeth, Demirjian 8-teeth, and Willems’ methods, and the skeletal age by applying Tanner Whitehouse-3-RUS (TW3-RUS) and Greulich & Pyle methods. Two machine learning models, Random Forest and Boosted, were created and trained on 70 % of all age estimates and then tested on the remaining 30 %. The results obtained by the AI for the test sample were compared to the performance of each original method. Results: The model built using Boosted machine learning for estimated age performed better than Random Forest, with a mean prediction range of 1087 days (±1.48 years), including 95 % of the estimated sample. This error is smaller than that of the traditional methods based only on tooth mineralization or wrist bone maturation. Conclusions: Application of the AI-assisted approach to a sample of wrist-hand and dental radiographs taken on the same date from the same subject demonstrates that combining multiple age estimates based on skeletal and dental methods improves the accuracy and reliability of the final age assessment.