A hybrid approach to predicting and classifying dental impaction: integrating regularized regression and XG boost...

Area of Science:

• Dental medicine
• Healthcare analytics
• Artificial intelligence in dentistry

Background:

• Dental impaction, a tooth alignment issue, presents a clinical challenge requiring advanced predictive modeling.
• Radiographic measurements like panoramic radiographs and CBCT are used for diagnosis, with AI enhancing prediction accuracy.

Purpose of the Study:

• To predict the eruption of mandibular third molars using a hybrid approach combining regularized regression and ensemble methods.
• To develop a robust clinical decision support system for dental practitioners by improving impaction prediction accuracy.

Main Methods:

• Quantitative, observational, cross-sectional retrospective study analyzing three parameters: 2nd molar to anterior border distance, 3rd molar mesiodistal width, and root apex to inferior border distance.
• Utilized ensemble learning algorithms integrated with regularized regression techniques for feature selection and predictive modeling.

Main Results:

• The hybrid approach achieved 78% accuracy in predicting dental impaction.
• Horizontal impaction showed the lowest space/width ratio (0.9267), indicating low eruption potential.
• Regularized logistic regression model attained 75% accuracy for classification and prediction.

Conclusions:

• The study identified key parameters (distance from 2nd molar to anterior ramus border and 3rd molar mesiodistal width) for predicting the space/width ratio.
• Advanced modeling techniques and data quality enhancement are crucial for improving predictive capabilities in dental impaction.
• Findings support optimized treatment planning and reduced potential complications for dental practitioners.