Tooth Numbering And Classification On Bitewing Radiographs: An Artificial Intelligence Pilot Study

Summary

This study investigates the use of deep learning techniques for the automated identification and enumeration of permanent teeth in bitewing radiographs. Accurate tooth numbering is critical for clinical decision-making, as it can save time and reduce the risk of errors.

Objective

The aim of this study is to evaluate the effectiveness of deep learning models in detecting and accurately numbering permanent teeth in bitewing radiographs. Integration of this technology into dental practice can accelerate diagnostic processes and improve accuracy.

Methods

• Data Collection: A total of 1248 bitewing radiographs were annotated using the CranioCatch labeling program developed in Eskişehir, Turkey.
• Deep Learning Model: A YOLOv5-based deep learning model was developed and trained to identify and enumerate permanent teeth.
• Training and Testing: The dataset was divided into training (n = 1000, 80%), validation (n = 124, 10%), and test (n = 124, 10%) sets to assess the model’s accuracy and generalizability. A 3 × 3 clash operation was applied to the images to enhance the clarity of labeled regions.
• Performance Metrics: Model performance was evaluated using F1 score, sensitivity, and precision, and predictions were compared with expert human evaluations.

Results

1. Model Performance:
   • The YOLOv5-based model achieved an F1 score of 0.9913, sensitivity of 0.9954, and precision of 0.9873 on the test dataset, demonstrating high accuracy and reliability in identifying permanent teeth.
2. Efficiency and Speed:
   • The model provided results much faster than manual evaluations, offering a significant advantage in clinical settings.
3. Clinical Applications:
   • AI-assisted automatic tooth identification and numbering systems can support dentists in clinical decision-making and improve diagnostic accuracy.

Conclusion

This study demonstrates that deep learning-based automated evaluation models are effective tools for identifying and numbering permanent teeth in bitewing radiographs. Integration of this technology into dental practice can enhance the speed and accuracy of diagnosis, ultimately improving clinical efficiency.