Reliability of a Deep Learning Model in Predicting Permanent Maxillary Canine Eruption for Preventive Orthodontics: A...

Area of Science:

• Orthodontics and Dental Imaging
• Artificial Intelligence in Healthcare
• Preventive Dentistry

Background:

• Impacted maxillary canines require careful management for aesthetic and functional outcomes.
• Deep learning models offer advanced diagnostic capabilities for therapeutic planning.
• Early prediction of canine eruption is crucial for preventive orthodontic strategies.

Purpose of the Study:

• To compare the diagnostic accuracy of a deep learning model against human experts in predicting permanent maxillary canine eruption.
• To evaluate the efficiency of a deep learning system in the preventive management of unerupted maxillary permanent canines (UPMCs).

Main Methods:

• A retrospective analysis of 2230 panoramic radiographs from patients aged 9-14 years.
• Classification of radiographs into three sectors using the modified Ericson and Kurol system.
• Development and training of a DenseNet121 Convolutional Neural Network (CNN) for predicting UPMC sectors.
• Evaluation of the AI model using accuracy, precision, recall, and F1 score.

Main Results:

• The deep learning model achieved high accuracy and precision across all three sectors for predicting UPMC eruption.
• For example, Sector 1 (right side) showed 99.55% accuracy and 100% precision; Sector 3 (left side) achieved 99.46% accuracy and 99.72% precision.
• The AI model demonstrated superior efficiency in landmark identification and parameter generation.

Conclusions:

• A deep learning-based system significantly enhances the efficiency of diagnostic processes in interceptive/preventive orthodontics.
• This AI tool reduces the time and human resources needed for evaluating unerupted maxillary permanent canines.
• The findings support the integration of AI for improved orthodontic diagnostics and treatment planning.