Skip to content

Latest commit

 

History

History
113 lines (86 loc) · 4.52 KB

LAG.md

File metadata and controls

113 lines (86 loc) · 4.52 KB

LAG

Dataset Information

The LAG dataset is a collection of 11,760 fundus images specifically designed for the early detection and study of glaucoma. The images in this dataset are derived from routine fundus examinations and are categorized into two sample types: 4,878 images as suspicious-glaucoma samples, and 6,882 images as non-glaucoma samples. Each image is accompanied by detailed diagnostic labels, where "0" indicates a non-glaucoma sample, and "1" indicates a suspicious-glaucoma sample.

Additionally, 5,824 images in the LAG dataset have further been annotated with attention maps. These annotations highlight areas of high risk identified by medical experts in the fundus images. Among these annotated images, 2,392 are positive for glaucoma, while the remaining 3,432 are classified as non-glaucoma.

The LAG dataset not only includes a large number of expert-annotated suspicious and negative samples but also features eye-tracking-related attention map annotations. This provides robust support for deep learning-based fundus image analysis and automated diagnostics, aiding the advancement of early screening and prevention systems for glaucoma, ultimately improving the management of ophthalmic diseases worldwide.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Number of Categories Data Volume File Format
2D Fundus Classification Retina Head and Neck 2 11760 JPG

Resolution Details

Dataset Statistics size
min (112, 112)
median (500, 500)
max (500, 500)

Label Information Statistics

Description Non-Glaucoma Suspicious-Glaucoma
Number of Images 6882 4878
Percentage (%) 58.5 41.5
Resolution (cm²) (112, 112) (112, 112)
Medium Resolution (500, 500) (500, 500)
High Resolution (500, 500) (500, 500)

Visualization

Non-glaucoma.

Suspected glaucoma.

Focus Area Map.

File Structure

LAG/
├── non_glaucoma/
│   ├── attention_map/
│   │   ├── image1.jpg
│   │   ├── image2.jpg
│   │   └── ...
│   ├── image/
│   │   ├── image1.jpg
│   │   ├── image2.jpg
│   │   └── ...
│   └── label/
│       ├── 0001.txt
│       ├── 0002.txt
│       └── ...
├── suspicious_glaucoma/
│   ├── attention_map/
│   │   ├── image5.jpg
│   │   ├── image8.jpg
│   │   └── ...
│   ├── image/
│   │   ├── image5.jpg
│   │   ├── image8.jpg
│   │   └── ...
│   └── label/
│       ├── 0005.txt
│       ├── 0008.txt
│       └── ...

Source Information

Official Website: https://github.com/smilell/AG-CNN?tab=readme-ov-file

Download Link: https://github.com/smilell/AG-CNN?tab=readme-ov-file

Article Address: https://arxiv.org/abs/1903.10831

Publication Date: 2019-03

Citation

@InProceedings{Li_2019_CVPR,
author = {Li, Liu and Xu, Mai and Wang, Xiaofei and Jiang, Lai and Liu, Hanruo},
title = {Attention Based Glaucoma Detection: A Large-Scale Database and CNN Model},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}

Original introduction article is here.