Automated Detection and Classification Of Tooth Types And Dental Anomalies In Panoramic Radiographs

Abstract

This paper proposes an automated deep learning framework for detecting and classifying dental anomalies from Orthopantomogram (OPG) X-rays using a Transformer-enhanced YOLOv10 model. The system simultaneously identifies four tooth types (incisors, canines, premolars, molars) and two distinct anomaly categories (caries and periapical lesions) in a single inference pass. A Flask-based web application is integrated for real-time prediction and visualization, enabling immediate clinical deployment without specialized infrastructure. The model achieves an overall mAP@0.5 of 92.0%, outperforming baseline YOLO variants by up to 7.8 percentage points. The system supports clinical decision-making in dentistry by reducing manual analysis errors, improving diagnostic throughput, and providing accessible AI-assisted dental screening in resource-constrained environments.