Dental Panoramic Radiography Dataset for Pixel level Semantic Segmentation of Teeth

This dataset consists of panoramic dental X ray (OPG) images collected from a clinical imaging facility in Sylhet, Bangladesh. The dataset contains 329 panoramic radiographs obtained from 329 individual patients and is intended for research purposes in dental image analysis and deep learning applications, particularly semantic segmentation of teeth. The data collection was carried out in coordination with the clinical imaging facility, which allowed the collection of panoramic dental radiographs for research purposes on the condition that no patient personal information would be accessed, recorded, or disclosed at any stage of the study. All images were fully anonymized prior to dataset creation to ensure patient confidentiality and ethical compliance.

The dataset includes both male and female subjects across different age groups, covering both adult and pediatric cases. It reflects a wide range of dental conditions commonly observed in clinical practice, including complete and incomplete dentition, missing teeth, restorations, impacted teeth, and other developmental variations. This diversity makes the dataset suitable for developing and evaluating robust deep learning models capable of handling real world clinical variability. The majority of samples correspond to adult patients.

The images were acquired from panoramic radiographs using smartphone photography under controlled conditions with a high resolution mobile device to preserve anatomical details while maintaining consistency and patient privacy. Although controlled conditions were maintained during acquisition, the dataset still exhibits natural variability in brightness, contrast, sharpness, and overall image quality due to real world clinical imaging environments.

All images were manually annotated at the pixel level for semantic segmentation of teeth. The annotation process was performed under the supervision of a professional dental expert to ensure anatomical correctness and clinical reliability. The segmentation masks were carefully reviewed through multiple quality assurance steps to maintain consistency and reduce annotation errors across the dataset.

This dataset is intended to support research in dental image analysis, particularly semantic tooth segmentation using deep learning and computer vision techniques. It is also applicable to automated dental assessment, clinical decision support systems, and AI based healthcare applications. The realistic variability and clinical diversity present in the dataset make it suitable for evaluating the robustness and generalization capability of segmentation models under real world conditions.

本数据集包含从孟加拉国锡尔赫特的一家临床影像机构采集的全景牙科X射线（OPG）图像。该数据集共包含329张来自329名独立患者的全景牙科X光片，旨在用于牙科图像分析与深度学习应用相关研究，尤其面向牙齿语义分割任务。本次数据采集与该临床影像机构协同开展，机构同意为研究用途采集全景牙科X光片，前提是在研究的任何阶段均不得访问、记录或披露患者个人信息。在构建数据集前，所有图像已完成完全匿名化处理，以保障患者隐私与研究伦理合规性。

本数据集涵盖不同年龄段的男性与女性受试者，包含成人及儿童病例。其囊括了临床实践中常见的多种牙科病症，包括完整牙列、不全牙列、牙齿缺失、修复体、阻生齿及其他发育变异情况。这种多样性使得该数据集适用于开发与评估能够应对真实临床变异性的鲁棒深度学习模型。数据集的绝大多数样本来自成人患者。

图像采集采用高分辨率移动设备，在受控条件下通过智能手机拍摄全景X光片，以保留解剖细节，同时确保采集一致性与患者隐私。尽管采集过程维持了受控条件，但受真实临床影像环境影响，数据集仍在亮度、对比度、清晰度及整体图像质量上呈现自然变异性。

所有图像均针对牙齿语义分割任务完成了像素级手动标注。标注流程在专业牙科专家的监督下开展，以确保解剖学正确性与临床可靠性。分割掩码经过多轮质量核查步骤的严格审核，以保障数据集内标注的一致性并降低标注误差。

本数据集旨在支撑牙科图像分析相关研究，尤其是基于深度学习与计算机视觉技术的牙齿语义分割研究。其同样可应用于自动化牙科评估、临床决策支持系统及基于人工智能的医疗保健应用。数据集所具备的真实变异性与临床多样性，使其适用于评估分割模型在真实场景下的鲁棒性与泛化能力。