Mule_Fundus_Image

Fundus images are captured by handheld fundus camera. Currently, only part of fundus image are annotated with forvea location. Compared to desktop fundus camera, handheld camera suffers more unstable capturing and uncontrolled envrionments. We are now focusing on the enhancement algorithm for this kind of low-quality fundus image. The Mule dataset is a fundus image dataset with reflection of retinal nerve fiber layer (RNFL). It is composed of 538 color fundus images captured using a handheld fundus camera (Mocular CFC-X camera with field of view equals to 45 degrees). Each image is preprocessed into 486*486 pixels and data desensitization is performed. The Mule dataset is a subset of a large-scale database that was collected from a fundus screening program at the Community Hospitals in the Lin'an District of Hangzhou, China. Patient information such as the age, gender, and ethnicity are not provided眼底图像（fundus image）由手持式眼底相机（handheld fundus camera）采集。目前仅对部分眼底图像标注了黄斑凹（fovea）的位置。与台式眼底相机相比，手持式眼底相机的成像过程更不稳定，且拍摄环境更难管控。本研究团队目前正针对这类低质量眼底图像研发对应的图像增强算法。 Mule数据集是一类带有视网膜神经纤维层（retinal nerve fiber layer, RNFL）反射特征的眼底图像数据集。该数据集包含538张由手持式眼底相机（Mocular CFC-X，视场角为45°）采集的彩色眼底图像。所有图像均被预处理至486×486像素，并完成了数据脱敏（data desensitization）操作。Mule数据集源自中国杭州市临安区社区医院眼底筛查项目所搭建的大型数据库，为该数据库的一个子集。该数据集未提供患者的年龄、性别、种族等个人身份信息。