Accurate prediction of three-dimensional tooth movement remains a major challenge in orthodontic treatment planning, with conventional methods showing error rates of 30-50% for complex movements. Therefore, it is of interest to develop and evaluate an artificial intelligence model for predicting orthodontic tooth movement using digital treatment records and intraoral scan data. A deep learning framework combining convolutional neural networks and recurrent neural networks was trained on 4,218 orthodontic cases comprising 892,476 individual tooth movement records across multiple treatment stages. The model achieved an overall prediction accuracy of 91.3%, with a mean absolute error of 0.24 mm for linear movement and 1.87°C for angular movement, significantly outperforming traditional prediction approaches (p = 0.001). Thus, we show that AI-based tooth movement prediction can enhance orthodontic treatment planning accuracy, reduce chairside time and improve overall clinical outcomes.
Keywords: Artificial Intelligence; deep learning; digital orthodontics; orthodontic tooth movement; treatment prediction.
© 2026 Biomedical Informatics.