MMDental - A multimodal dataset of tooth CBCT images with expert medical records

In the rapidly evolving field of dental intelligent healthcare, where Artificial Intelligence (AI) plays a pivotal role, the demand for multimodal datasets is critical. Existing public datasets are primarily composed of single-modal data, predominantly dental radiographs or scans, which limits the development of AI-driven applications for intelligent dental treatment. In this paper, we collect a MultiModal Dental (MMDental) dataset to address this gap. MMDental comprises data from 660 patients, including 3D Cone-beam Computed Tomography (CBCT) images and corresponding detailed expert medical records with initial diagnoses and follow-up documentation. All CBCT scans are conducted under the guidance of professional physicians, and all patient records are reviewed by senior doctors. To the best of our knowledge, this is the first and largest dataset containing 3D CBCT images of teeth with corresponding medical records. Furthermore, we provide a comprehensive analysis of the dataset by exploring patient demographics, prevalence of various dental conditions, and the disease distribution across age groups. We believe this work will be beneficial for further advancements in dental intelligent treatment.在人工智能（Artificial Intelligence, AI）发挥关键作用的牙科智能医疗快速发展领域中，多模态数据集的需求至关重要。现有的公开数据集主要由单模态数据构成，以牙科X光片或扫描影像为主，这限制了面向智能牙科诊疗的人工智能驱动应用的发展。在本研究中，我们采集并构建了多模态牙科（MultiModal Dental, MMDental）数据集以填补这一空白。该数据集涵盖660名患者的数据，包括3D锥形束计算机断层扫描（3D Cone-beam Computed Tomography, CBCT）影像，以及包含初始诊断与随访记录的详细专家医疗档案。所有CBCT扫描均在专业医师指导下完成，所有患者档案均经资深医师审核。据我们所知，本数据集是首个且规模最大的包含牙齿3D CBCT影像及对应医疗档案的数据集。此外，我们对该数据集开展了全面分析，涵盖患者人口统计学特征、各类牙科疾病患病率以及不同年龄组的疾病分布情况。我们相信本研究将有助于推动牙科智能诊疗的进一步发展。