Home

Awesome

RIM-ONE DL

RIM-ONE for Deep Learning

RIM-ONE DL is a unified retinal image database for assessing glaucoma using Deep Learning. The full paper is available in this publication of the Image Analysis and Stereology journal: https://www.ias-iss.org/ojs/IAS/article/view/2346

This repository hosts the RIM-ONE DL image dataset and the related data and tools, which consists of:

Using the database

Data included in this database can only be used for research and educational purposes, free of charge and without requesting permission to the authors. Copy, redistribution, and any unauthorized commercial use are prohibited.

In order to use this database and have comparable results among different publications, please use the original partitions described below for training and testing purposes. Moreover, use only the data included in RIM-ONE DL, do not add more images from different databases to train your model or tune your algorithm.

If you use RIM-ONE DL in your work, please cite the following publication:

FUMERO BATISTA, Francisco José et al. RIM-ONE DL: A Unified Retinal Image Database for Assessing Glaucoma Using Deep Learning. Image Analysis & Stereology, v. 39, n. 3, p. 161-167, nov. 2020. ISSN 1854-5165. Available at: https://www.ias-iss.org/ojs/IAS/article/view/2346. doi: https://doi.org/10.5566/ias.2346.

BibTeX format:

@article{RIMONEDLImageAnalStereol2346,
	author = {Francisco José Fumero Batista and Tinguaro Diaz-Aleman and Jose Sigut and Silvia Alayon and Rafael Arnay and Denisse Angel-Pereira},
	title = {RIM-ONE DL: A Unified Retinal Image Database for Assessing Glaucoma Using Deep Learning},
	journal = {Image Analysis & Stereology},
	volume = {39},
	number = {3},
	year = {2020},
	keywords = {Convolutional Neural Networks; Deep Learning; Glaucoma Assessment; RIM-ONE},
	issn = {1854-5165},
	pages = {161--167},
	doi = {10.5566/ias.2346},
	url = {https://www.ias-iss.org/ojs/IAS/article/view/2346}
}

Images

The RIM-ONE DL image dataset consists of 313 retinographies from normal subjects and 172 retinographies from patients with glaucoma. These images were captured in three Spanish hospitals: Hospital Universitario de Canarias (HUC), in Tenerife, Hospital Universitario Miguel Servet (HUMS), in Zaragoza, and Hospital Clínico Universitario San Carlos (HCSC), in Madrid.

This dataset has been divided into training and test sets, with two variants:

Download images

The images can be downloaded as a ZIP file from the following link: https://bit.ly/rim-one-dl-images

This ZIP file contains the two variants of the dataset (partitioned randomly and by hospital).

Reference segmentations of the optic disc and cup

All the images of RIM-ONE DL include a manual segmentation of the disc and cup performed by an expert in glaucoma.

These manual segmentations were carried out using DCSeg.

Download reference segmentations

The reference segmentations can be downloaded from the following link: https://bit.ly/rim-one-dl-reference-segmentations

This ZIP file contains the segmentation of the disc and cup for each image in PNG format and in DCSeg TXT format.

Convert DCSeg TXT files to NumPy or PNG

If you would like to use the DCSeg TXT files, in this repository, we have included some Python tools to convert the masks generated by DCSeg to NumPy arrays and PNG.

The main file is generate_dir_binary_masks.py. This script reads the RIM-ONE DL images and reference segmentations directories and converts the TXT masks to PNG, storing the PNG masks in the given output directory. To run the script, open it with your editor of choice and adjust the directories and the bottom of the file to point to your actual directories.

The previous file makes use of the dcseg_to_binary_mask function, which takes the paths to an image and a DCSeg mask and returns a NumPy 2D boolean mask.

CNN Weights

Together with the publication of the RIM-ONE DL database, we described an evaluation benchmark with different models of well-known convolutional neural networks, which includes: Xception, VGG16, VGG19, ResNet50, InceptionV3, InceptionResNetV2, MobileNet, DenseNet121, NASNetMobile and MobileNetV2.

These networks were trained using the Keras Deep Learning Framework. In every case, the size of the input layer was set to 224x224x3, and a GlobalAveragePooling2D layer was added to the convolutional base, followed by a fully-connected output layer with two outputs, using SoftMax to distinguish between the Normal and Glaucoma classes.

The following tables show the results achieved by these networks to classify RIM-ONE DL images between Normal and Glaucoma classes.

Evaluation of the different networks using the random test set:

NetworkAUCSeAcc.
VGG190.98671.00000.9315
VGG160.98340.96150.9247
Xception0.97710.98080.9178
ResNet500.97550.98080.9110
MobileNetV20.97380.94230.9041
DenseNet0.97260.96150.9041
MobileNet0.97120.96150.9315
InceptionResNetV20.96850.98080.9110
InceptionV30.95970.94230.8904
NASNetMobile0.92900.92310.7534

Evaluation of the different networks using the test set partitioned by hospital:

NetworkAUCSeAcc.
VGG190.92720.87500.8563
VGG160.91770.82140.8506
InceptionV30.90150.75000.8046
Xception0.89820.75000.7989
DenseNet0.89190.71430.7816
MobileNet0.89120.75000.8276
ResNet500.88550.73210.8333
InceptionResNetV20.83960.6250.7644
NASNetMobile0.79690.60710.7989
MobileNetV20.77650.44640.5287

Metrics used in the tables above:

Submit your own results to this page

If you use RIM-ONE DL to train your own method or tune your algorithm and want your results to be published in this page, please contact us with the following info:

Download CNN Weights

The weights of the CNNs used in the publication can be downloaded from: https://bit.ly/rim-one-dl-cnn-weights

This ZIP file contains the weights in h5 format of each trained network for the two variants of the dataset, i.e. partitioned randomly and partitioned by hospital.