A novel collaborative learning model for mixed dentition and fillings segmentation in panoramic radiographs

Area of Science:

• Artificial Intelligence in Dentistry
• Machine Learning for Medical Imaging

Background:

• Accurate interpretation of dental radiographs is crucial for diagnosis and treatment planning.
• Human error and image quality issues can lead to misinterpretations of dental radiographs.
• Existing AI models lack the ability to simultaneously differentiate tooth types and detect fillings.

Purpose of the Study:

• To develop a novel collaborative learning model for analyzing panoramic radiographs.
• To simultaneously identify and differentiate primary and permanent teeth.
• To detect dental fillings in panoramic radiographs.

Main Methods:

• Developed two high-performance classifiers: one for tooth segmentation (differentiating primary and permanent teeth) and one for dental filling detection.
• Utilized publicly accessible and University of Missouri-Kansas City dental panoramic radiographic images.
• Created a novel collaborative learning method integrating the two classifiers to enhance performance.

Main Results:

• The tooth segmentation classifier achieved a mean average precision (mAP) of 95.32% and F-1 score of 92.50%.
• The dental filling detection classifier achieved an mAP of 91.53% and F-1 score of 91.00%.
• The collaborative learning model demonstrated enhanced performance with an mAP of 94.09% and F-1 score of 93.41%.

Conclusions:

• The developed AI model effectively identifies and differentiates primary and permanent teeth and detects associated dental fillings.
• This model represents an improvement over existing machine learning approaches for panoramic radiograph analysis.
• The AI tool can support dentists' interpretations, patient communication, and dental student education.