The model of "Attention-based CNN for Glaucoma Detection (AG-CNN)", which has been published as "Attention Based Glaucoma Detection: A Large-scale Database with a CNN Model".
The LAG database contains 11,760 fundus images corresponding to 4,878 suspecious and 6,882 negative glaucoma samples. All the samples are labeled with the diagnosis results (0 refers to negative glaucoma and 1 refers to suspecious glaucoma). 5,824 fundus images are further labeled with attention regions based on an alternative method for eye tracking, in which 2,392 are positive glaucoma and the rest 3,432 are negative glaucoma.
An example of capturing fixations of an ophthalmologist in glaucoma diagnosis.
Some samples from our LAG database.
We have uploaded the first part of our LAG database at Dropbox under request. Please contact us for the passport.
Note that the LAG database should ONLY be used for academic purpose and other usage is refused. Also, it is NOT allowed to re-upload the LAG database on the internet.
Our work is conducted according to the tenets of Helsinki Declaration. As the retrospective nature and fully anonymized usage of color retinal fundus images, we are exempted by the medical ethics committee to inform the patients.
The conference version of our work has been published in CVPR2019, one can cite with the Bibtex code:
@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}
}
If any question, please contact [email protected] or [email protected].