Detection of periodontal bone loss patterns and furcation defects from panoramic radiographs using deep learning...

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Periodontal bone loss is a key indicator of gum disease severity.
• Panoramic radiographs are commonly used for dental assessments.
• Accurate detection of bone loss patterns is crucial for effective treatment planning.

Purpose of the Study:

• To develop a deep learning algorithm for interpreting panoramic radiographs.
• To evaluate the algorithm's performance in detecting periodontal bone loss and patterns.
• To assess the AI system's utility in identifying furcation defects.

Main Methods:

• A Convolutional Neural Network (CNN) with U-Net architecture was developed.
• 1121 panoramic radiographs were analyzed.
• Bone losses (total alveolar, interdental, furcation) and patterns (horizontal, vertical) were meticulously labeled using segmentation.

Main Results:

• The AI system achieved high diagnostic performance, with an Area Under the Curve (AUC) of 0.951 for total alveolar bone loss.
• Performance varied by defect type, with the lowest AUC of 0.733 for vertical bone loss.
• High sensitivity and precision were observed for total alveolar bone loss detection.

Conclusions:

• AI systems, specifically CNN algorithms, show significant promise for identifying periodontal bone loss and furcation defects from radiographs.
• This technology can potentially automate the assessment of periodontal disease severity.
• The findings suggest future applications in treatment planning and analysis of various dental radiographs.