A Composite Retinal Fundus and OCT Dataset to Grade Macular and Glaucomatous Disorders

Abstract

Retinopathy represents a group of retinal diseases that causes severe visual impairments and even blindness. Many researchers have publicly released datasets containing fundus or optical coherence tomography (OCT) scans to screen retinal diseases like macular edema (ME) and age-related macular degeneration (AMD). These datasets also contain clinical markings to analyze the retinal layers and retinal lesions within normal and abnormal pathologies. However, to the best of our knowledge, no dataset provides the clinically graded fundus and OCT images reflecting the geographic AMD, neovascular AMD, acute central serous retinopathy (CSR), chronic CSR, centrally involved DME (ci-DME), and glaucomatous pathologies. Furthermore, the majority of the publicly available OCT datasets are acquired through Spectralis Machines, which limits the thorough evaluation of autonomous frameworks to screen retinal pathologies irrespective of the scanner specifications. To overcome these challenges, we present a novel dataset containing composite fundus and OCT scans of each patient, along with detailed annotations for extracting the retinal layers and retinal lesions. Also, contrary to its competitors, the proposed dataset is acquired through Topcon 3D OCT 2000 machine that can be utilized for training (or evaluating) any autonomous frameworks to give the lesion-aware screening and severity grading of the above-mentioned retinal diseases as per the clinical standards. Moreover, in this paper, we are also releasing the retinal annotation software alongside the proposed dataset. This software can help clinicians in quickly marking both fundus and OCT scans, which can be saved later on in any image format. Overall, the proposed dataset contains 9,268 OCT scans and 180 fundus scans from 105 subjects depicting healthy, ci-DME, geographic AMD, neovascular AMD, acute CSR, chronic CSR, and glaucomic pathologies.