BACKGROUND: Accurate detection and numbering of teeth on panoramic radiographs are essential for diagnosis and treatment planning. Manual interpretation is time-consuming and prone to variability, particularly in mixed dentition. Deep learning offers potential for reliable automated analysis. AIM AND OBJECTIVES: To develop and evaluate a deep learning-based model for automated detection and Fédération Dentaire Internationale (FDI)-based identification of teeth in mixed dentition panoramic radiographs. Objectives included tooth segmentation, detection, and automatic numbering. MATERIALS AND METHODS: A retrospective dataset of 670 pediatric panoramic radiographs (6-12 years) was used. Images were annotated using semantic and instance segmentation and divided into training (80%), validation (10%), and test (10%) sets. A YOLOv11 instance segmentation model was trained using transfer learning. Performance was assessed using precision, recall, F1-score, and mean average precision (mAP). RESULTS: The model demonstrated excellent performance with precision, recall, and F1-score of approximately 99.8%. mAP values reached ~99-100% at IoU 0.5 and ~98-99% at IoU 0.5-0.95. Slightly reduced recall was observed for third molars. CONCLUSION: The proposed model provides highly accurate automated tooth detection and FDI-based numbering in mixed dentition, with strong potential for integration into clinical dental workflows.
Singh et al. (Sun,) studied this question.