sites/RIPS/DatasetRIPS

RIPS: Retinal Images for Pigment Signs

Introduction

The RIPS dataset has been created in order to evaluate and compare methods for the segmentation of pigment signs (PSs) in retinal images. It provides data and ground truth, so that everyone can test their own algorithm. On this page, instructions can be found on downloading the dataset.

Data description

The RIPS dataset includes 120 retinal images acquired from four patients, where 99 contain PSs and in 21 PSs are absent. Each patient underwent three different acquisition sessions. During each session, five retinal images per eye have been captured covering different regions of the fundus. The images have been acquired using the digital retinal camera Canon CR4-45NM (Canon UK, Reigate, UK). The resolution of the images is 1440×2160. The time lapse between two consecutive sessions may vary from one to six months, while the total period between the first and the last session always exceeds one year. Unless the limited number of patients, this dataset is particularly challenging, since images show high variability regarding color balancing, contrast and sharpness/focus, also for the same patient. Moreover, the dataset provides images with and without PSs. Two experts, with a different level of confidence in the field of ophthalmology, have been involved in the ground truth generation process. Thus, RIPS dataset includes two sets (G1 and G2) of manually segmented images to be used as ground truths for performance evaluations. Moreover, for each image, a mask image is provided that delineates the Field of View (FoV).

Data organization

The dataset contains four directories:

  • Original that contains the input color retinal image;
  • G1 and G2 that contain the ground truths;
  • FoV that contains the masks that delineate the field of view of the retina.

For each directory, a folder for each subject is present: subject1, subject2, subject3 and subject4.

In each subject folder there are three folders: session1, session2 and session3.

For each session folder, we have two sub-folders: left and right, indicating left and right eye, respectively.

Finally, each left and right folder contains five images.

Small versions of all the images, together with their ground truths, can be viewed for subject: Subject 1, Subject 2, Subject 3, Subject 4.

In the tables, the text color of the names of the images without PSs is red.

Download

The agreement for the use of the dataset is available here. The compiled module should be sent to Nadia Brancati or Diego Gragnaniello, which will provide the instructions for the download.

Using the dataset

If you use the RIPS dataset to test and report results in any publication, we request you to do so by citing this publication:

  • N. Brancati, M. Frucci, D. Gragnaniello, D. Riccio, V. Di Iorio, L. Di Perna, F. Simonelli “Learning-based approach to segment pigment signs in fundus images for Retinitis Pigmentosa analysis”, Neurocomputing (submitted)

In addition, we appreciate to hear about any publications that use the RIPS dataset. Feedback on the dataset and this website is also welcome. The persons to contact are Nadia Brancati or Diego Gragnaniello.

Back To Top