眼底筛查AI特征数据集

该数据集源自爱尔大数据中心2020年8月1日至2023年12月15日的眼底影像，数据范围涵盖10个省份、17家医院，经专家标注后形成眼底筛查AI特征数据集，数据规模：23195条。该数据集以临床应用为导向，基于 22 种眼病相关指标（含 18 种眼病）及 76 种病灶指标特征进行标注，涵盖糖尿病性视网膜病变、高血压性视网膜病变、青光眼等常见病症，形成AI训练样本，供爱尔自研模型：眼底图像质量评估模型与眼底筛查 AI 模型进行训练，通过眼底影像眼病及病灶特征识别和提取，输出包含眼底影像描述、AI 疑似适应症、初步建议及科普内容的眼底筛查分析报告，辅助临床实现眼底病变早期识别与高效诊断，有效赋能眼底基层筛查场景。​

This dataset is derived from fundus images acquired by Aier Big Data Center between August 1, 2020 and December 15, 2023. The data covers fundus images from 17 hospitals across 10 provinces. After expert annotation, it is formulated into a fundus screening AI feature dataset with a total of 23,195 samples. Clinically oriented, this dataset is annotated based on 22 ophthalmology-related indicators (including 18 types of eye diseases) and 76 lesion characteristic indicators. It covers common ocular disorders such as diabetic retinopathy, hypertensive retinopathy, glaucoma and other related conditions, and serves as AI training samples for two self-developed models by Aier: the fundus image quality assessment model and the fundus screening AI model. By identifying and extracting features of eye diseases and lesions from fundus images, these models can generate fundus screening analysis reports containing fundus image descriptions, AI-suspected indications, preliminary recommendations and popular science content. This assists clinicians in early identification and efficient diagnosis of fundus diseases, and effectively empowers grassroots fundus screening scenarios.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于眼底疾病筛查的AI特征数据集，包含23195条眼底影像数据，由爱尔眼科医院集团通过公开收集和专家标注构建。数据集支持18种眼病和76种病灶的识别，应用于AI辅助诊断模型训练，旨在提升眼底病变早期筛查效率和远程诊疗能力。