Dental age estimation by comparing Demirjian's method and machine learning in Southeast Brazilian youth.

This study evaluated the applicability of combining Demirjian's method with machine learning algorithms to estimate the chronological age of children and adolescents from southeastern Brazil, using dental development stages as predictive variables. A retrospective study was conducted using 610 digital panoramic radiographs of children and adolescents. Demirjian's method was applied to classify the permanent mandibular teeth into eight developmental stages. Eight machine learning models-Linear Regression, Gradient Boosting Regressor, K-Nearest Neighbors Regressor, Support Vector Regression, Multilayer Perceptron Regressor, Decision Tree Regressor, Random Forest Regressor, and AdaBoost Regressor-were trained and evaluated using five-fold cross-validation. Model accuracy was compared to the traditional method using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R²). Paired Student's t-tests were used to compare actual chronological age with predicted age estimates, and bootstrapping with 1,000 iterations was performed to calculate 95% confidence intervals (CI95%). Machine learning-based models achieved predictive errors of less than 1.5 years. The Gradient Boosting and Random Forest models demonstrated the highest performance, with an MAE of 0.75 (95% CI: [0.66-0.85]) and an RMSE of 0.92 (95% CI: [0.81-1.05]), representing a 44.03% reduction in MAE and a 43.56% reduction in RMSE compared to Demirjian's method (MAE = 1.34, RMSE = 1.63). Integrating machine learning with Demirjian's method improved the accuracy of dental age estimation, reducing errors and enhancing the reliability of the approach. The application of artificial intelligence reduces the mean absolute error of the dental age estimation method. This approach can optimize diagnoses and assist in both clinical and forensic settings, providing a more precise and adaptable tool for diverse populations.