Contact Endoscopy – Narrow Band Imaging (CE-NBI) Data Set for Laryngeal Lesion Assessment

The endoscopic examination of subepithelial vascular variations of vocal folds can provide complementary diagnostic information for clinicians regarding the development of benign and malignant laryngeal lesions. As one novel technique, Contact Endoscopy combined with Narrow Band Imaging (CE-NBI) can provide real-time and enhanced visualization of these vascular structures. Several studies have addressed the concern of subjective evaluation of CE-NBI images, resulting in the development of multiple computer-based solutions. We introduce the CE-NBI data set, the first publicly available data set with enhanced and magnified visualization of vocal fold subepithelial blood vessels. It comprises 11144 images of 210 adult patients with benign and malignant lesions in the vocal fold. Image annotations include as following for all images of every patient: Diagnosed laryngeal histopathology label. Lesion type benign-malignant label. Leukoplakia diagnosis label. The dataset consists of two main categories: benign and malignant images. In each category, the images of every patient are ordered according to the laryngeal histopathology class. Additionally, one Excel file is provided to map the image files of each patient to three image labels and image dimensions. This data has successfully been used to perform clinical evaluations as well as design and develop multiple Machine Learning (ML)-based algorithms for laryngeal cancer assessment.声带黏膜下血管变异的内镜检查，可为临床医师提供喉良恶性病变发生的辅助诊断依据。作为一项新兴内镜技术，窄带成像联合接触式内镜（Contact Endoscopy combined with Narrow Band Imaging，CE-NBI）可对上述血管结构实现实时增强可视化。已有多项研究针对CE-NBI图像的主观评估难题展开攻关，由此催生了多种基于计算机的辅助解决方案。本研究公开的CE-NBI数据集，是首个可实现声带黏膜下血管增强放大可视化的公开数据集。该数据集涵盖210名成年声带良恶性病变患者的11144张内镜图像。针对每位患者的所有图像，均标注以下三类信息：确诊喉组织病理学标签、病变良恶性分类标签、白斑病诊断标签。本数据集分为良性图像与恶性图像两大类别，每类别下的患者图像均按照喉组织病理学分级进行排序。此外，数据集附带一份Excel文件，用于将每位患者的图像文件与上述三类图像标签及图像尺寸信息进行关联映射。该数据集已成功应用于临床评估工作，同时被用于设计开发多种用于喉癌评估的机器学习（Machine Learning，ML）算法。