An Intraoperative Hyperspectral Imaging Dataset with Parathyroid-Focused Annotations During Thyroidectomy

The annotation scheme is centred on the parathyroid gland. When available, neighbouring tissues and key anatomical structures, including thyroid gland, trachea, fat tissue, lymph node, and muscle, are also provided. Unlabelled pixels are treated as ignore regions by default.All data have been de-identified before release.

The dataset does not contain patient names, identity numbers, medical record numbers, hospital numbers, contact information, local acquisition-folder tokens, or other direct personal identifiers. Because the complete dataset contains human-derived intraoperative surgical-scene images and health-related medical context, access to the full dataset should be managed under a controlled-access mechanism. Data users should comply with the Data Usage Agreement, should not attempt to re-identify participants, should not redistribute the primary data without permission, and should use the data only under appropriate ethical and data-security conditions.This dataset can be used for spectral analysis of parathyroid glands and surrounding tissues, parathyroid localization and identification, validation of medical hyperspectral image reading and preprocessing workflows, reuse of task-oriented annotations, patient-level leakage-aware benchmark construction, and the development and validation of intraoperative intelligent tissue-recognition algorithms.