Classification of caries in third molars on panoramic radiographs using deep learning

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Dental caries detection on panoramic radiographs (PRs) is crucial for diagnosis.
• Accurate classification of carious lesions in third molars can be challenging.
• Deep learning offers potential for improving diagnostic accuracy in dentistry.

Purpose of the Study:

• To evaluate the classification accuracy of dental caries on PRs using a deep learning algorithm.
• To assess the performance of the MobileNet V2 convolutional neural network (CNN) for caries detection in third molars.
• To determine the feasibility of AI-driven analysis for third molar assessments.

Main Methods:

• A CNN, specifically MobileNet V2, was trained on 400 cropped panoramic images of third molars.
• The trained MobileNet V2 model was tested on a separate set of 100 cropped PRs.
• Classification accuracy, sensitivity, specificity, and area-under-the-curve (AUC) were calculated.

Main Results:

• The MobileNet V2 algorithm achieved a classification accuracy of 0.87.
• The model demonstrated a sensitivity of 0.86 and a specificity of 0.88.
• An AUC of 0.90 was recorded, indicating strong performance in classifying carious lesions.

Conclusions:

• Deep learning, specifically the MobileNet V2 algorithm, shows high accuracy in classifying dental caries on PRs of third molars.
• The findings support the development of automated AI-based systems for third molar assessment.
• This technology has the potential to enhance diagnostic efficiency and accuracy in dental practice.