Descriptor: Periodontal Bone-Loss Assessment of Panoramic Radiograph Dataset (BoneLoss-PAN769)

Abstract

Large expert-curated datasets are essential for reliable artificial intelligence (AI) in dental imaging, but manual annotation remains slow and costly. Detailed crown annotations are particularly valuable for computer-aided analysis of natural, restored, and prosthetic crowns, as well as for downstream geometric assessment of alveolar structures. We present BoneLoss-PAN769, an expert-guided dataset of 769 panoramic radiographs annotated through a hybrid human-AI workflow. Expert-provided cementoenamel-junction (CEJ) and alveolar-ridge landmarks, acquired in a customized Picture Archive and Communication System (PACS) interface, are combined with automated tooth instance segmentation to derive toothwise crown regions of interest, stored together with the alveolar-ridge polylines in a single common objects in context (COCO)-compatible JSON file. We demonstrate the utility of the dataset by training per-class YOLOv11x-Seg instance-segmentation models, chosen for their accuracy-speed tradeoff and for producing object-level masks that facilitate downstream geometric analyses. On a held-out patientwise test set, the models achieved a mask mean Average Precision (mAP@0.5:0.95) of 70.4% for crowns and 89.6% for the alveolar ridge; fivefold cross-validation with 95% confidence intervals and a paired statistical test confirm the results and a small but significant benefit of data augmentation. We also developed an end-to-end application for automated periodontal bone-loss quantification based on the crown and alveolar-ridge segmentations. In a perceptual study with three expert evaluators, the system outputs received a favorable accuracy rating, with perceived accuracy greater than 70% in 95.1% of the evaluated cases. The dataset adheres to the FAIR principles and is publicly available, providing a high-precision resource for periodontal bone-loss research and broader dental AI applications. IEEE SOCIETY/COUNCIL IEEE Engineering in Medicine and Biology Society (EMBS) DATA TYPE/LOCATION Panoramic dental radiographs; School of Dentistry of Ribeirão Preto, University of São Paulo, Brazil DATA DOI/PID 10.13026/s6wh-nq14