Loading BCONZ...
Ophthalmology datasets
Ophthalmology datasets for AMD, diabetic retinopathy and glaucoma research
Eye disease research depends on long follow-up and, increasingly, on imaging. BCONZ provides governed, de-identified US ophthalmology datasets for age-related macular degeneration, diabetic retinopathy and glaucoma — for real-world evidence, disease progression studies and the development and validation of retinal AI.
Not a free download. These are governed, research-grade datasets licensed for a defined purpose after a feasibility review — built for studies and products that public datasets cannot support.
Research uses
What ophthalmology datasets are used for
Retinal AI development and external validation
Train or externally validate diabetic retinopathy, AMD or glaucoma models on real-world data from outside the development cohort.
Disease progression
Study progression from early to advanced AMD, or visual field and structural change in glaucoma, over longitudinal follow-up.
Treatment patterns and outcomes
Real-world use and outcomes of anti-VEGF therapy, glaucoma medication and surgery.
Screening programmes
Evidence for diabetic retinopathy screening pathways and referral outcomes.
How access works
- Tell us your research question through the research data request form.
- We confirm feasibility: whether a suitable cohort exists, its size, follow-up and available data types.
- Governance review and a data use agreement define the permitted purpose.
- The de-identified dataset is prepared, quality-reviewed and delivered securely.
Hold ophthalmology data yourself? A healthcare data readiness assessment shows what it can support, and DIA shows who has published a need for it.
Frequently asked questions
Ophthalmology datasets: common questions
Do the ophthalmology datasets include retinal images (OCT or fundus)?
Imaging availability varies by partner and is confirmed during feasibility, together with how images link to the clinical record.
What are the ophthalmology patient numbers based on?
Figures are approximate and rounded down. They count patients in de-identified US EMR/EHR data with a recorded diagnosis in the area; a patient can appear in more than one indication. The cohort for a specific study is confirmed during feasibility.
How are these ophthalmology datasets different from public datasets on Kaggle or GitHub?
Public datasets are valuable for learning and benchmarking, but they are usually small, single-source, fixed snapshots with limited clinical context and no route to more data. BCONZ datasets are research-grade, de-identified and licensed through governed partnerships, with longitudinal records and a feasibility step to confirm the cohort fits your study before any agreement.
Is this a free dataset download?
No. Access is licensed for a defined research or development purpose under a data use agreement, after a feasibility review. Tell us what you need through the research data request form.
Where does the data come from?
From US healthcare data partners, and from partner networks in other regions where a study needs them. Data is de-identified before research use and stays governed by the partner's agreement.