Akademik Veri Yönetim Sistemi
International Congress of DentoMaxilloFacial Radiology, London, İngiltere, 24 Haziran 2025, ss.26, (Özet Bildiri)
Abstract:
Materials and Methods: A dataset of 2,999 panoramic radiographs with 40,383 restorations was collected from the radiological archive of Eskisehir Osmangazi University Faculty of Dentistry. Multi-class dental restorations including crowns, fillings, pontics, root canal-filling, dental implants, implant-supported crowns, and posts were annotated using CranioCatch annotation tool (Eskisehir, Turkiye). The dataset was split into training (80%), validation (10%), and test (10%) subsets. The YOLOv8m segmentation model was trained for 500 epochs, and performance was evaluated using precision, recall, and F1 score.
Results: The YOLOv8m based multi-class segmentation model achieved high accuracy in detecting and classifying various dental restorations on panoramic radiographs. It demonstrated a sensitivity of 99.15%, recall of 98.05%, and an F1 score of 98.60%, with 3,976 true positives, 34 false positives, and 79 false negatives. The model successfully identified multiple restoration classes including crowns, fillings, pontics, root canal-filling, dental implants, implant-supported crowns, and posts.
Conclusion: The YOLOv8m-based artificial intelligence model effectively automates the detection of multi-class dental restorations on radiographic images, significantly improving diagnostic accuracy and consistency. This tool has substantial potential for clinical application in evaluating restorations and assisting in treatment planning. Future research using larger datasets and additional restoration categories may further enhance the model’s performance and broaden its clinical relevance.