(PO31) Evaluation of the Accuracy of an Artificial Intelligence Algorithm in Diagnosing Periapical Pathosis on Periapical Radiographs

Abstract: The objective of this study was to verify the diagnostic performance of an artificial intelligence algorithm for the diagnosis of periapical pathosis on periapical radiographs, using an AI architecture based on a deep learning-based convolutional neural network (CNN). In this study, 150 periapical images of patients with periapical pathosis were used, out of 210 periapical radiographs. Patient radiographs were reviewed from the EPIC and MiPacs systems of the University of Mississippi Medical Center, from January 2020 to September 2023. Teeth associated with periapical pathosis were determined by an experienced oral and maxillofacial radiologist and an endodontist. The deep learning-based CNN model was used for the detection of periapical pathosis. The AI model was trained using the CNN-based architecture with 500 epochs and a learning rate of 0.01. The Pearson Product Moment was used for interobserver agreement, and a confusion matrix was used for statistical evaluation. The Artificial Intelligence (CranioCatch, Turkey) system was used for manual segmentation and analysis. As per our initial assessment, the deep learning CNN algorithm was successful in the detection of periapical pathosis on periapical radiographs. The accuracy of correctly detecting a periapical lesion was 87.3%. There was no significant difference between the two measurement methods (p > .05). The difference in periapical pathosis detection performed by specialists and by the AI algorithm was not statistically significant. The performance of the CNN architecture in the detection of periapical pathosis on periapical radiographs has the potential for clinical use in assisting endodontists.