Long-Term Predictive Modelling of the Craniofacial Complex Using Machine Learning on 2D Cephalometric Radiographs.

Area of Science:

• Orthodontics and craniofacial development
• Machine learning applications in healthcare
• Predictive modeling in biology

Background:

• Long-term prediction of craniofacial growth is crucial for orthodontic treatment planning.
• Accurate forecasting of skeletal and dental changes aids in personalized treatment strategies.
• Existing methods may lack precision in predicting complex growth trajectories.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting long-term craniofacial growth.
• To assess the accuracy of different machine learning algorithms in forecasting skeletal and dental relationships.
• To identify key predictive factors for craniofacial development.

Main Methods:

• Utilized cephalometric data from 301 subjects (pre-pubertal T1 and post-pubertal T2).
• Trained three machine learning models (Lasso, Random Forest, SVR) on 240 subjects.
• Validated model performance using mean absolute error (MAE), intraclass correlation coefficients (ICCs), and clinical thresholds (2 mm or 2°).

Main Results:

• Machine learning models achieved clinically acceptable prediction margins (2 mm or 2°) for several craniofacial measurements.
• Prediction accuracy was higher for skeletal relationships (e.g., maxilla to cranial base angle at 80%) than dental relationships.
• Pre-pubertal measurements and sex were identified as the most significant predictors of post-pubertal craniofacial characteristics.

Conclusions:

• Machine learning models effectively predict key post-pubertal craniofacial skeletal and dental relationships.
• The models offer clinically relevant accuracy for predicting changes in angles and positions over an 8-year period.
• Early cephalometric data and sex are vital for accurate long-term craniofacial growth prediction.