DentVoxel: a fully annotated dental CBCT dataset with 38 instance anatomical structures

Cone-beam computed tomography (CBCT) is widely used in clinical dental diagnosis and treatment planning by providing high-resolution spatial visualization of teeth, alveolar bone, and surrounding anatomical structures. High quality instance level annotated datasets are essential for advancing artificial intelligence driven dental imaging analysis. Therefore, this study introduces DentVoxel, a densely annotated CBCT dataset comprising 100 large field-of-view scans with 38 instance level labels, including the maxilla, mandible, bilateral maxillary sinuses, bilateral mandibular canals, and 32 individual teeth labeled according to the FDI numbering system. All images and labels are provided in NIfTI format and are publicly available, providing a high quality open resource for dental anatomical modeling, three-dimensional instance segmentation, and AI based CBCT analysis.