UK Biobank retinal imaging grading: methods, baseline characteristics and findings for common ocular diseases.

Abstract

Purpose : The UK Biobank (UKBB) study is a large, prospective cohort study including adults registered with the National Health Service. The aims of the present study were to describe the grading methods and baseline characteristics for participants who underwent retinal imaging, and to characterise individuals with retinal features suggestive of age-related macular degeneration (AMD), glaucoma and diabetic retinopathy (DR).

Methods : Non-mydriatic colour fundus photographs and macular optical coherence tomography (OCT) scans were acquired using a Topcon 3D OCT-1000 Mark II system (Topcon, Japan). Grading was performed by trained and certified graders and quality assured by clinicians of the Network of Ophthalmic Reading Centres UK (NetwORC UK – Belfast, Liverpool, and Moorfields Ophthalmic Reading Centres). Study-specific grading forms were created to capture retinal features including those associated with any AMD (at least one definite drusen on either imaging modality), glaucoma (based on colour fundus photographs only; ≥0.7 cup-disc ratio, ≥0.2 cup-disc ratio difference between the eyes, abnormal disc features (notching, inferior rim thinning or both) and/or disc haemorrhage in either eye) and DR (microaneurysms with or without other characteristic features of DR). Suspected cases of these conditions were compared against the information provided by UKBB self-reported (verbal interview), linked hospital episode statistics and available primary care records.

Results : Among 68,517 UKBB participants who underwent retinal imaging, the median age at imaging was 58 years (interquartile range 51-64), 45.7% were men and 90.6% were of white ethnicity. Altogether, 64,266 (93.8%) participants had gradable colour fundus photographs and 68,465 (99.9%) had gradable OCT scans in at least one eye. Retinal features suggestive of any AMD, glaucoma and DR were identified in 15,178 (22.2%), 2,183 (3.2%) and 265 (0.4%) participants, of whom 125 (0.8%), 218 (10.0%) and 132 (49.8%) respectively had a recorded diagnosis.

Conclusions : The outcomes of grading of baseline images represents a rich resource that will aid detailed phenotyping of UKBB subjects and yields diagnostic information not captured by self-report and healthcare data linkage. Judicious use of the image analysis information generated will accelerate research across a range of areas including genetics and artificial intelligence.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.