Optimization technique combined with deep learning method for teeth recognition in dental panoramic radiographs.

Area of Science:

• Computer-aided diagnostics
• Artificial intelligence in dentistry
• Medical imaging analysis

Background:

• Computer-assisted analysis of dental radiographs is crucial for reducing human error and enhancing diagnostic accuracy.
• Automated systems can improve efficiency and reduce diagnosis time in dental care.

Purpose of the Study:

• To propose an automatic teeth recognition model for dental radiographs.
• To enhance the accuracy and reliability of dental diagnostic tools.

Main Methods:

• Utilized a Faster R-CNN technique based on residual networks for automatic teeth recognition.
• Implemented a candidate optimization technique to refine detection results based on positional relationships and confidence scores.
• Conducted tenfold cross-validation to assess model robustness and feasibility.

Main Results:

• Achieved high mean Average Precision (mAP) scores of 0.974 (ResNet-50) and 0.981 (ResNet-101) using Faster R-CNN.
• The optimization technique improved the F1 score from 0.978 to 0.982 for ResNet-101.
• Tenfold cross-validation demonstrated robust performance with an average F1 score exceeding 0.970.

Conclusions:

• The proposed model accurately recognizes teeth in dental radiographs.
• The automated system demonstrates high accuracy and robustness, making it a reliable tool for dental professionals.
• This technology has the potential to significantly assist in dental diagnostics and improve overall dental care efficiency.