Prediction of Pulpal Sequelae in Cracked Teeth with Reversible Pulpitis using Machine Learning Models.

Area of Science:

• Dentistry
• Machine Learning
• Biostatistics

Background:

• Cracked teeth with reversible pulpitis pose a diagnostic challenge regarding the need for root canal treatment (RCT).
• Retained pulp vitality in cracked teeth correlates with better survival rates, while RCT can negatively impact outcomes.
• Accurate prediction of pulp survival is crucial for timely and appropriate endodontic intervention.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting pulp survival in cracked teeth with reversible pulpitis.
• To investigate patient- and tooth-related variables associated with treatment outcomes.
• To enhance diagnostic precision in managing cracked teeth requiring endodontic evaluation.

Main Methods:

• Analysis of data from 593 cracked teeth across 569 patients.
• Application of Logistic Regression, Gaussian Processes, Random Forests, and Gradient Boosting models.
• Utilized 10-fold stratified nested cross-validation for model performance estimation and hyperparameter optimization.

Main Results:

• Logistic Regression model achieved the highest Area Under the Curve (AUC) of 0.64 and F1-score of 0.60.
• Models demonstrated strong Positive Predictive Value (PPV) ranging from 0.74 to 0.77, indicating effective identification of teeth requiring RCT.
• Older age and the presence of preoperative restorations were significant predictors, suggesting these patients are less likely to need RCT.

Conclusions:

• Machine learning models achieved a predictive accuracy of 74-77% for pulp survival in cracked teeth.
• These ML models can significantly improve diagnostic accuracy for endodontic decision-making.
• The findings support the use of ML in clinical practice for managing cracked teeth with reversible pulpitis.