Debrecen 糖尿病视网膜病变数据集

该数据集包含从梅西多图像集中提取的特征，用于预测图像是否包含糖尿病视网膜病变的迹象。所有特征代表检测到的病变、解剖部位的描述性特征或图像级描述符。Balint Antal、Andras Hajdu：基于集成的糖尿病视网膜病变自动筛查系统，基于知识的系统60（2014年4月），20-27中描述了图像分析和特征提取的基本方法以及我们的分类技术。图像集（Mesidor）可在[Web link].上找到。 Attribute Information: 0）质量评估的二元结果。0=质量差1=质量足够。 1） 预筛选的二元结果，其中1表示严重视网膜异常，0表示缺乏。 2-7）MA检测结果。每个特征值代表在置信水平α=0.5，分别为1。 8-15）包含与2-7）相同的渗出物信息。然而由于渗出物由一组点表示，而不是由构成病变的像素数表示，因此通过将病变数除以ROI直径来对这些特征进行归一化，以补偿不同的图像大小。 16） 黄斑中心和视盘中心的欧氏距离提供了重要信息 关于病人的情况。该特征也会随着ROI的直径而标准化。 17） 视盘的直径。 18） 基于AM/FM分类的二进制结果。 19） 类标签。1=包含DR迹象（梅西多1、2、3类的累积标签），0=无DR迹象。 Relevant Papers: Provide references to papers that have cited this data set in the past (if any). Citation Request: Please cite the following paper: Balint Antal, Andras Hajdu: An ensemble-based system for automatic screening of diabetic retinopathy, Knowledge-based Systems 60 (April 2014), 20-27. The dataset is based on features extracted from the Messidor image dataset: [Web link].

This dataset contains features extracted from the Messidor image dataset, which is used to predict whether an image contains signs of diabetic retinopathy (DR). All features represent either detected lesions, descriptive features of anatomical locations, or image-level descriptors. The fundamental methods of image analysis and feature extraction, as well as our classification techniques, were described in Balint Antal, Andras Hajdu: An ensemble-based system for automatic screening of diabetic retinopathy, Knowledge-Based Systems 60 (April 2014), 20-27. The Messidor image dataset is available at [Web link].

Attribute Information:
1) Binary result of quality assessment: 0 = poor quality, 1 = sufficient quality.
2) Binary result of pre-screening, where 1 indicates severe retinal abnormalities and 0 indicates their absence.
3–7) Results of microaneurysm (MA) detection. Each feature value corresponds to a confidence level of α=0.5, with 1 indicating a positive detection.
8–15) Contain exudate information identical to that of features 2–7. However, since exudates are represented as a set of points rather than the number of pixels constituting the lesions, these features are normalized by dividing the number of lesions by the diameter of the ROI to compensate for varying image sizes.
16) The Euclidean distance between the center of the macula and the center of the optic disc provides critical information about the patient's condition. This feature is also normalized by the diameter of the ROI.
17) Diameter of the optic disc.
18) Binary result based on AM/FM classification.
19) Class label: 1 = contains signs of DR (cumulative label for Messidor classes 1, 2, and 3), 0 = no signs of DR.

Relevant Papers: Provide references to papers that have cited this dataset in the past (if any).
Citation Request: Please cite the following paper: Balint Antal, Andras Hajdu: An ensemble-based system for automatic screening of diabetic retinopathy, Knowledge-Based Systems 60 (April 2014), 20-27. The dataset is based on features extracted from the Messidor image dataset: [Web link].

搜集汇总

背景与挑战

背景概述

Debrecen糖尿病视网膜病变数据集包含从梅西多图像集中提取的特征，用于预测糖尿病视网膜病变的迹象。数据集提供了详细的属性信息，包括质量评估、病变检测结果和图像描述符，适用于医学图像分析和自动筛查系统的研究。