Evaluation of artificial intelligence systems in brand and model identification of dental implants

Objective: Dental implants have become one of the most popular and accepted rehabilitation methods for missing teeth. Dental implants, which contain many parts, have a unique design and dentists need to be able to identify the manufacturer of the implants in order to provide effective treatment. The aim of our study is to evaluate the effectiveness of artificial intelligence systems in identifying the make and model of various dental implants. Method: In this study, panoramic radiographs from patients who underwent dental implant treatment at Akdeniz University Faculty of Dentistry between 2020 and 2023 were retrospectively collected. A total of 523 panoramic radiographs including 1907 implants from seven implant models of five manufacturers were included in the dataset. Data augmentation was applied to this dataset with brightness and rotation modifications. The new dataset was divided into 80% training, 10% validation and 10% test data, and the implants were labeled using the LabelImg program. YOLOv5, YOLOv7 and YOLOv8 algorithms, which are artificial intelligence based object detection methods, were used to identify the implant systems. Precision, sensitivity, mAP, F1 scores and confusion matrices were used to evaluate the performance of the algorithms. Results: The precision, recall, mAP and F1 scores across all implants were 0.786, 0.648, 0.764 and 0.71 for YOLO v5; 0.81, 0.815, 0.855 and 0.81 for YOLO v7; and 0.89, 0.901, 0.955 and 0.89 for YOLOv8, respectively. The algorithms generally performed better for implants with a larger number in the dataset and different implant designs. Conclusion: Artificial Intelligence based object detection algorithms have performed adequately in identifying dental implants on panoramic radiographs, even on a relatively small dataset. As the algorithms were updated, their performance gradually improved. Artificial intelligence systems trained with larger datasets and different dental implants could help physicians to identify implants.