BIOMISA Retinal Image Database for Retinopathy

Retinopathy is a collective group of syndromes that damages the human retina due to increased fluid pressure or hyperglycemia. The major forms of retinopathy include macular edema (ME), central serous retinopathy (CSR), exudative and non-exudative age-related macular degeneration (AMD). Here we present a high-resolution retinal image database containing 64 fundus and 2497 OCT brightness scans (B-scans). The proposed dataset is first of its kind in providing detailed annotations of retinal hemorrhages, hard exudates, intra-retinal and sub-retinal fluids, drusen, retinal pigment epithelium (RPE) atrophy and cup to disc (CDR) ratios from both retinal fundus and OCT imagery.

Dataset Acquisition Details:

The dataset has been acquired from Armed Forces Institute of Ophthalmology (AFIO), Pakistan. The data acquisition phase was carried out under the strict observation of multiple expert ophthalmologists. Apart from this, the annotations were carried out individually by each ophthalmologist based on their knowledge and expertise, in an isolated zone. All the images are acquired using TOPCON 3D OCT 2000 machine. Fundus scans are centered on both macular and ocular region, with the dimensions of 2032 x 1934 pixels. OCT images are characterized as B-scans and C-scans where B-scans are acquired with the resolution 951 x 456 pixels and C-scans contains 128 frames where each frame has a resolution of 760 x 576 pixels. Fundus and OCT B-scans are stored in JPEG uncompressed images while OCT C-scans are stored in WMV format with 15 frames per second.

Dataset Composition Details:

The dataset contains 64 fundus scans (20 maculae centered and 44 OD centered) and 2497 OCT B-scans (2453 are of macular region and 44 are of OD region). Out of these 2497 B-scans, 2432 B-scans are embedded into 19 C-scans of 128 frames. The total patient count for macular region is 59 (14 AMD subjects, 13 ME subjects and 32 healthy subjects) and for ocular region, it is 44 (26 glaucomic and 18 healthy subjects). Furthermore, all the scans are arranged patient wise in the dataset and for each patient, the scans are distinguished w.r.t left and right eye.

Citation:

Please cite following article in order to use this dataset for research purposes. Please note that the commercial use of this dataset is prohibited.

  • Taimur Hassan, M. Usman Akram, M. Furqan Masood, Ubaid Ullah Yasin, “BIOMISA Retinal Image Database for Macular and Ocular Syndromes”, Image Analysis and Recognition, Springer, Portugal, June 2018.

Request for Dataset: