Deep learning-based prediction of indication for cracked tooth extraction using panoramic radiography

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Cracked tooth syndrome is a common dental issue often requiring extraction.
• Early diagnosis of cracked teeth is crucial for effective treatment planning.
• Panoramic radiography is a widely used imaging technique in dentistry.

Purpose of the Study:

• To assess the feasibility of using deep learning models to predict cracked tooth extraction indications.
• To evaluate the diagnostic performance of AI algorithms on panoramic radiographs for cracked teeth.
• To determine the accuracy and reliability of deep learning in identifying teeth requiring extraction due to cracks.

Main Methods:

• A deep learning model was trained and tested using panoramic radiographs of 418 teeth (209 normal, 209 cracked).
• Three distinct deep learning architectures (InceptionV3, ResNet50, EfficientNetB0) were evaluated for cracked tooth diagnosis.
• The model underwent fivefold cross-validation with a 3:1:1 ratio for training, validation, and testing data.

Main Results:

• The deep learning models achieved sensitivity rates between 90.43% and 94.26%.
• Specificity ranged from 52.63% to 60.77%, with overall accuracy between 72.01% and 75.84%.
• F1 scores for the models were reported between 76.36% and 79.00%.

Conclusions:

• Deep learning models show potential in predicting cracked tooth extraction indications from panoramic radiography.
• While sensitivity is high, further improvements in specificity are needed for clinical application.
• AI-assisted analysis of panoramic radiographs could aid in the early detection of cracked teeth.