Automatic feature segmentation in dental panoramic radiographs

Abstract

Objective

The purpose of our study is to verify the diagnostic performance of an artificial intelligence (AI) system for the automatic detection of teeth, caries, implants, restorations, and fixed prosthesis on panoramic radiography.

Methodology

Panoramic radiographs were obtained from the EPIC and MiPacs systems of the University of Mississippi Medical Center, spanning from June 2022 to May 2023. A total of one thousand panoramic radiographs of adults were used to identify teeth, caries, implants, restorations, and fixed prostheses. The study included images from 580 patients. The identification and detection of teeth, caries, implants, restorations, and fixed prostheses were then independently determined by 2 oral and maxillofacial radiologists. The convolutional neural network−based architecture was analyzed for detecting panoramic findings. The artificial intelligence system (Velmeni Inc.) was used for analysis to determine whether the panoramic findings could be detected.

Results

The convolutional neural network system successfully detected teeth, caries, implants, restorations, and fixed prostheses on panoramic radiography. The AI system was able to detect findings in 567 out of a total of 580 panoramic radiographs, with a reliability of correctly detecting panoramic findings at 97.75%.

Conclusion

The detection of teeth and periapical pathosis performed by oral radiologists and by AI systems were comparable with each other. AI systems developed on the basis of on deep-learning methods can be useful for detecting teeth, caries, implants, restorations, and fixed prosthesis on panoramic images for clinical applications.