An artificial intelligence proposal to automatic teeth detection and numbering in dental bite-wing radiographs

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Radiological examination is crucial in dental practice, with intraoral imaging frequently employed.
• Accurate tooth numbering on radiographs is a time-consuming routine task for dentists.

Purpose of the Study:

• To develop an automated system for detecting and numbering teeth in bitewing radiographs.
• To utilize a faster Region-based Convolutional Neural Network (R-CNN) for efficient tooth identification.

Main Methods:

• A dataset of 1125 bitewing radiographs from Ordu University Faculty of Dentistry (2018-2019) was analyzed.
• A faster R-CNN model was implemented for object identification of teeth.
• Model performance was evaluated using a confusion matrix, F1 score, precision, and sensitivity.

Main Results:

• The deep Convolutional Neural Network (CNN) system correctly numbered 697 out of 715 teeth in 109 test images.
• Achieved high performance metrics: F1 score of 0.9515, precision of 0.9293, and sensitivity of 0.9748.
• The CranioCatch system demonstrated significant accuracy in tooth detection and numbering.

Conclusions:

• A CNN-based approach shows significant potential for analyzing bitewing images.
• Automated tooth detection and numbering can substantially reduce the time dentists spend on dental charting.
• This technology offers a promising solution for streamlining dental radiographic analysis.