Nuclear Cataract Database for Biomedical and Machine Learning Applications

The dataset collected for multi-class nuclear cataract classification consists of 1437 labeled images. The images were obtained from a digital Topcon DC-3 slit-lamp camera with a resolution of 3264 x 2448 RGB; they were focused with a Topcon SL-D7 slit-lamp, a slit angle of 45º and 135º for each eye, a 6 mm x 0.5 mm aperture and maximum luminous intensity. The number of volunteers for the multi-class dataset was 68 women and 47 men, with at least three images per right and left eye. The expert ophthalmologists from the Conde Valenciana Institute from Mexico classified this dataset according to the LOCS III Protocol.本数据集用于多分类核性白内障（nuclear cataract）分类任务，共包含1437张标注图像。图像由拓普康DC-3型数码裂隙灯相机（Topcon DC-3 slit-lamp camera）采集，分辨率为3264×2448的RGB格式；拍摄时采用拓普康SL-D7型裂隙灯（Topcon SL-D7 slit-lamp）完成对焦，每只受试眼分别采用45°与135°的裂隙角度，光圈设置为6 mm×0.5 mm，并使用最大发光强度。本数据集的志愿者共计68名女性与47名男性，每只右眼与左眼至少采集3张图像。来自墨西哥孔德瓦伦西亚研究所（Conde Valenciana Institute）的资深眼科医师依据LOCS III分级方案（LOCS III Protocol）对该数据集完成了分类标注。