Akademik Veri Yönetim Sistemi
2025 10th International Conference on Computer Science and Engineering (UBMK), İstanbul, Türkiye, 15 Eylül - 21 Aralık 2025, cilt.1, sa.1, ss.1-6, (Tam Metin Bildiri)
In this study, various YOLO-based deep learning models were employed for the detection of dental findings in panoramic dental radiographs, and their performances were comparatively evaluated. Due to irregular class distribution and labeling problems in the 31-class Dental X-Ray Panoramic Dataset consisting of 13,376 images, data filtering and optimization processes were performed. The number of classes was reduced to 4 based on clinical importance (Filling, Implant, Root Canal Treatment, Impacted Tooth), creating an optimized dataset of 1,105 images, and various YOLO model architectures were tested and their performances were compared on this dataset. CLAHE technique was applied to improve image quality, and different sized versions of YOLOv8 and YOLOv11 models were tested using 5-fold cross-validation. Results show that YOLOv11 series models exhibit more consistent performance. The best performance value was obtained as 88% mAP@50 with the YOLOv11-large-seg model. In the class-based analysis results, the implant class was observed to be the most successful class with 98.5% mAP@50. This study provides important findings for the development of artificial intelligence-supported automatic diagnosis systems in dental radiology.