DentalOPG-1550: A Panoramic Dental X-Ray Image Dataset for Dental Disease Research

Abstract

DentalOPG-1550 is a clinical panoramic dental radiography dataset developed from routine diagnostic orthopantomogram (OPG) examinations. The original collection contains 1,550 panoramic dental radiographs representing real-world variations in image brightness, contrast, sharpness, patient positioning, anatomical overlap, dental morphology, and disease presentation. These characteristics make the dataset suitable for the development and evaluation of artificial-intelligence methods for dental image analysis. This release includes an object-detection-ready annotated subset containing 1,525 panoramic radiographs. The annotations cover 14 dental conditions and findings: BDR, Bridge, Cap, Caries, Distoangular Impaction, Filling, Horizontal Impaction, Missing Teeth, Mixed Dentition, Normal Teeth, Periodontitis, Root Canal Treatment, Unerupted Wisdom Teeth, and Apical Abscess. The annotated dataset contains 14,497 object instances distributed across 14 dental categories, with Normal Teeth representing the largest class (7,280 instances), followed by Caries (1,578), Missing Teeth (1,547), Root Canal Treatment (894), BDR (810), Filling (738), Cap (636), Unerupted Wisdom Teeth (316), Horizontal Impaction (277), Distoangular Impaction (118), Apical Abscess (98), Mixed Dentition (95), Bridge (56), and Periodontitis (54). The annotated dataset is provided in YOLOv8 format and is divided into training, validation, and testing subsets containing 1,068, 303, and 154 images, respectively. These subsets contain 10,141, 2,934, and 1,422 annotated objects, respectively. During preprocessing, the images were auto-oriented and resized to 640 × 640 pixels using stretch resizing. The exported images are stored in JPEG format with RGB color representation. No image augmentation was applied in this release. The annotations were reviewed and verified for annotation consistency and clinical relevance by Sujon Debnath, Medical Technologist, Institute of Health Technology (IHT), Bogura, Bangladesh. Since a panoramic radiograph may contain multiple teeth and dental findings, a single image may contain several bounding boxes belonging to the same or different categories. DentalOPG-1550 can support research in dental condition recognition, multi-class object detection, deep learning, computer-aided diagnosis, medical image analysis, model benchmarking, and explainable artificial intelligence (XAI).