Akademik Veri Yönetim Sistemi
1st International Conference on Emerging Technologies and Engineering Systems, ICETES 2026, Hybrid, Amman, Ürdün, 7 - 09 Nisan 2026, ss.76-81, (Tam Metin Bildiri)
The automatic detection of dental caries in radiographic images through deep learning-based segmentation is considered a significant contribution to the progress of digital dentistry. In this study, a systematic comparison of three different deep learning models-FCN-ResNet50, DeepLabV3, and YOLOv8-was performed using a dataset of 175 panoramic radiographs (PRs). During the process, labeling reliability and data quality were prioritized over the total volume of the dataset. Each model was evaluated under identical conditions using four standard metrics: IoU, Dice coefficient, Precision, and Recall. According to the experimental results, the most successful performance was achieved by YOLOv8, which reached an IoU of 0.8052 and a Dice score of 0.8912. While competitive results were also observed with the FCN-ResNet50 and DeepLabV3 models, particularly regarding precision, YOLOv8 demonstrated superior overall effectiveness. It is concluded that deep learning frameworks, especially YOLOv8, provide a robust solution for accurate caries segmentation.