AI-Powered Automated 3D Facial Analysis To Predict Mci Width

Appropriate width of maxillary central incisor (MCI) is essential in smile rehabilitation, here, we developed an objective and efficient AI-powered automated workflow to predict MCI width via 3D facial analysis. This retrospective study (ethics approval #2024055) analyzed 3D facial scans of 200 adults (18-30 years) with natural dentition from January 2023 to December 2024. Facial landmarks (inner/outer canthi, pupils, alar bases, oral commissures) were automatically detected using dlib’s facial landmark model. Post-processing included 3D coordinate mapping through triangulation alignment. MCI width was measured from CBCT. Data augmentation via GAN-generated synthetic 3D facial scans was implemented to improve model robustness. Five machine learning algorithms, Multilayer Perceptron (MLP), Gradient Boosting, Random Forest (RF), Decision Tree, Multiple Linear Regression (MLR) were used to predict width of MCI, and models’ performance was evaluated by R-square (R²), mean square error (MSE), and root mean square error (RMSE). MLP performed best (Rtest²=0.97, RMSE=0.09 mm), followed by Gradient Boosting (Rtest²=0.945, RMSE=0.124 mm). Tree-based models exhibited inferior accuracy (RF Rtest²=0.89, RMSE=0.17 mm; decision tree Rtest²=0.80, RMSE=0.24 mm). MLR showed relatively higher accuracy (Rtest²=0.86, RMSE=0.20 mm). This novel AI system achieves submillimeter accuracy in MCI width prediction from automated 3D facial analysis and effectively avoiding manual measurement errors. By establishing quantifiable craniofacial-dental correlations through machine learning analysis, we provide a feasible method in digital prosthodontics design and smile rehabilitation.