Evaluation of artificial intelligence for automatic tooth and periapical pathosis detection on panoramic radiography

Tamara Vujanovic,Rohan Jagtap

DOI: https://doi.org/10.1016/j.oooo.2022.10.016

2023-02-19

Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology

Abstract:To verify the diagnostic performance of a convolutional neural network (CNN) artificial intelligence system for automatic tooth and periapical pathosis detection on panoramic radiography. Patient radiographs were reviewed from EPIC and MiPacs systems at the University of Mississippi Medical Center from June of 2019 to May of 2021. In total, 2113 panoramic radiographs of adults were used to identify periapical pathosis. Images of 218 patients were included. The specific areas of the jaw and teeth associated with the periapical lesions were then determined by a human observer. The CNN-based architecture was analyzed for automatic teeth detection. It was then used to determine whether the periapical pathosis could be detected and if the lesion was detected, where it was located. A manualsegmentation and artificial intelligence (AI) (Diagnocat Inc., San Francisco, CA, USA) system was used for analysis. The CNN system was successful in detecting teeth and periapical pathoses. Only one tooth was incorrectly identified. The AI system was able to detect 203 of a total of 218 periapical lesions. The reliability of correctly detecting a periapical lesion was 93.1%. There was no significant difference between the two measurement methods (P > 0.05). Tooth and periapical pathosis detection performed by humans and by AI systems were comparable to each other. AI systems based on deep learning methods can be useful for detecting teeth and periapical pathosis on panoramic images for clinical application. Human/Animal subjects were used and this study was approved by an institutional ethics panel.

dentistry, oral surgery & medicine