SCOPE-HN: A Segmentation-based Collection of OroPharyngeal Structures Using Flexible Endoscopy for Head and Neck Cancers

Oropharyngeal cancer (OPC) is one of the few cancers with a rising incidence, driven primarily by increasing rates of human papillomavirus (HPV)-associated disease. Early detection and accurate delineation of OPC are critical for diagnosis, treatment planning and improving outcomes but remain challenging due to anatomical complexity and variability in clinical expertise. We present the SCOPE-HN dataset, a curated collection of annotated endoscopic images from 106 patients with histologically confirmed OPC, collected at Stanford between 2015 and 2023. The dataset consists of 942 RGB images extracted from diagnostic nasopharyngolaryngoscopy videos, with pixel-level annotations performed by expert Head and Neck surgeons. Annotations include 12 semantic classes representing tumor, normal anatomical structures, and common endoscopic artifacts. The SCOPE-HN dataset is publicly available at <a href='https://scope-hn.stanford.edu'>scope-hn.stanford.edu</a> or <a href='https://bit.ly/SCOPE-HN'>bit.ly/SCOPE-HN</a> under the Stanford University Dataset Research Use Agreement license.