Predicting restoration failures in primary and permanent teeth - A machine learning approach.

Area of Science:

• Dental informatics
• Machine learning in healthcare
• Predictive modeling in dentistry

Background:

• Machine learning (ML) offers advanced capabilities for analyzing complex medical data.
• Predictive models are crucial for improving patient outcomes in dentistry.
• Accurate prediction of dental restoration failures is essential for effective treatment planning.

Purpose of the Study:

• To develop and evaluate machine learning (ML) predictive models for posterior dental restoration failures.
• To assess model performance in both primary and permanent teeth using clinical datasets.
• To explore the utility of ML in predicting restoration longevity.

Main Methods:

• Utilized data from two Randomized Controlled Trials (RCTs) for primary (CARDEC 3) and permanent teeth (CaCIA Trial).
• Developed models using Decision Tree, Random Forest, XGBoost, CatBoost, and Neural Network algorithms.
• Assessed model performance using accuracy, precision, recall, F1-score, ROC AUC, and SHAP plots for interpretability.

Main Results:

• Models for primary teeth showed acceptable performance with AUC values between 0.67-0.75.
• Models for permanent teeth demonstrated lower predictive ability, with AUC values ranging from 0.53-0.62.
• All models achieved a balanced trade-off between precision and recall for primary teeth.

Conclusions:

• ML models can effectively process complex, multi-dimensional data for predicting restoration longevity.
• Further validation with larger, diverse datasets is necessary for clinical implementation and improved accuracy.
• These models can potentially aid dentists in tailoring recall intervals and optimizing dental care allocation.