Machine learning to predict distal caries in mandibular second molars associated with impacted third molars

Area of Science:

• Dentistry
• Machine Learning in Healthcare
• Oral Health Research

Background:

• Impacted mandibular third molars (M3M) are frequently associated with distal caries on adjacent mandibular second molars (DCM2M).
• Early identification of DCM2M risk is crucial for effective clinical management and caries prevention strategies.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting DCM2M occurrence.
• To identify key predictive variables for DCM2M associated with M3Ms for clinical decision-making.

Main Methods:

• Analysis of 2642 mandibular second molars adjacent to impacted M3Ms.
• Development and validation of five ML models: logistic regression, random forest, support vector machine, artificial neural network, and extreme gradient boosting.
• Identification of significant predictive features including sex, age, cementoenamel junction contact point, M3M angulation, and classification systems (Winter's, Pell and Gregory).

Main Results:

• DCM2Ms were identified in 12.2% of the analyzed cases.
• ML models demonstrated superior performance compared to single predictors, with AUCs ranging from 0.88 to 0.89.
• Six key features were identified as significant predictors for DCM2M.

Conclusions:

• Validated ML models can effectively predict the risk of DCM2M in patients with impacted M3Ms.
• These models can aid clinicians in identifying high-risk individuals for targeted caries prevention and treatment.
• The findings support the integration of ML tools in dental practice for improved patient outcomes related to M3M complications.