Related Experiment Video
Updated: Jan 7, 2026
07:32
Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
1.7K
Machine Learning-Based Prediction Model for Alveolar Bone Defect Risk Following Orthodontic Treatment
Ailin Xu1, Sen Yang1, Shan Dong1
1Department of Orthodontics, Beijing Stomatological Hospital, Capital Medical University, Capital Medical University School of Stomatology, Beijing, China.
Orthodontics & Craniofacial Research
|December 25, 2025
Summary
This study developed an XGBoost machine learning model to predict alveolar bone defect risk after orthodontic treatment. The model accurately identifies high-risk patients, aiding personalized treatment planning.
Area of Science:
- Orthodontics and Dental Research
- Artificial Intelligence in Healthcare
- Biomedical Data Science
Background:
- Alveolar bone defects like fenestration and dehiscence are common complications of orthodontic treatment.
- These defects can negatively impact treatment outcomes and long-term oral health.
- Accurate risk assessment is crucial for preventing and managing these complications.
Purpose of the Study:
- To develop and validate a machine learning model for classifying the risk grade of alveolar bone defects post-orthodontic treatment.
- To identify key predictors of alveolar bone defect development.
- To provide a clinical decision-support tool for orthodontists.
Main Methods:
- Retrospective cohort study with 354 patients.
- Development of five machine learning models (XGBoost, SVM, MLP, Logistic Regression, Decision Tree).
- Feature selection using LASSO regression; performance evaluation using accuracy, AUC, precision, recall, F1-score.
- XGBoost model interpreted using SHAP analysis.
Main Results:
- LASSO regression identified 23 critical predictors; XGBoost model utilized 20 variables.
- The XGBoost model achieved superior performance with 0.9429 accuracy and 0.9955 AUC.
- Key predictors included pre-treatment bone defect score, basal bone arch width, and posterior mandible length.
- XGBoost significantly outperformed conventional methods.
Conclusions:
- An XGBoost model was successfully developed to predict alveolar bone defect risk.
- The model demonstrates high accuracy and outperforms existing methods.
- This tool can aid orthodontists in pre-treatment risk assessment and personalized treatment planning.
