Automated quality evaluation of dental panoramic radiographs using deep learning

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Panoramic radiographs are crucial for dental diagnosis but often suffer from quality issues affecting accuracy.
• Expert assessment of radiograph quality is time-consuming and inconsistent.
• Artificial intelligence (AI) offers a potential solution for automated, objective radiograph quality evaluation.

Purpose of the Study:

• To develop and evaluate a deep learning (DL)-based model for automated quality assessment of dental panoramic radiographs.
• To assess the model's performance across specific quality criteria including contrast, artifacts, coverage, and positioning.

Main Methods:

• A dataset of 1,000 panoramic radiographs was annotated by two dentists based on predefined quality criteria.
• Five YOLOv8 deep learning models were trained to classify specific quality aspects.
• Model performance was evaluated using accuracy metrics on a separate test set.

Main Results:

• The DL models achieved high accuracies for individual criteria: 97.9% for contrast/density, 87.2% for artifact detection, 79.3% for overall quality, 77.3% for positioning, and 74.1% for coverage.
• A model classifying images as clinically acceptable or unacceptable reached 81.4% average accuracy.

Conclusions:

• Deep learning-based automated quality assessment for panoramic radiographs is feasible.
• The developed model shows potential for integration into clinical workflows to enhance efficiency and consistency.
• This AI tool could also serve as an educational resource for dental students to improve radiographic techniques.