Creative Commons source reading · Machine learning for performance
Neural Network Technologies for Age Estimation in Children from Orthopantomograms (a Pilot Study)
Poletaeva MP; Vasilevsky YV; Valetov DK; Angelakopoulos N; Zolotenkova GV
Originally published in Sovrem Tekhnologii Med · 2026-04-30
From the article
Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.
Abstract
AbstractThe aim of the study was to investigate the potential of using artificial intelligence technologies for age estimation in children from dental radiographs.Materials and MethodsA retrospective study was conducted, analyzing orthopantomograms of 322 children (173 female, 149 male) aged 4–16 years. Fourteen permanent mandibular teeth were annotated on each radiograph. Neural network training was performed by splitting the data into training and test sets at a ratio of 80:20; 5-fold cross-validation was used. Age estimation was approached as a regression task. The neural network training and validation were conducted in Python using the PyTorch library. The accuracy of age prediction was assessed using the coefficient of determination (R2), mean squared error (MSE), and mean absolute error (MAE).ResultsThe study showed that the…
Results
During the search for the optimal training parameters, the following values were identified and established: learning rate — 0.001, input image edge size — 224 pixels, and batch size for training — 16. To build the ensemble, independent sequential training of the models was performed using the Adam method. Table 1 shows the MAE values for different ensemble sizes obtained during testing and optimal model selection. After that, the performance of the models within the ensemble was evaluated: predictions were obtained from each individual model on the test data and averaged with equal weights, which allowed to reduce prediction variance. Through cross-validation, the most optimal and acceptable training parameters were identified. Finally, an ensemble of five models was selected as the most accurate. When the ensemble size exceeded five…
Discussion
The majority of contemporary methods for dental age estimation involve a complex process of “manual” determination of a wide range of morphometric features, which are then compared to reference values stratified by sex and ethnicity, or involve measuring metric dental parameters followed by age calculation using regression equations. In international practice, the Demirjian’s and Cameriere’s methods are most commonly used for age estimation [3–5]. These methods have gained widespread acceptance across different countries due to their ease of use and acceptable results. However, “manual” methods have several limitations. The main one is connected with estimation difficulties caused by the task monotony and expert’s eyestrain, as well as lack of experience in analyzing radiographic images (orthopantomograms) and…
Editorial reading guide
- What data created and tested the model, and was evaluation independent?
- What is the cost of a false positive, false negative or out-of-distribution decision?
