Ophthalmology AI
Enabling AI to detect and monitor sight-threatening diseases at population scale.
Industry Challenge
Diabetic retinopathy, glaucoma, age-related macular degeneration, and other retinal conditions affect hundreds of millions globally, yet access to ophthalmology specialists is severely limited in low- and middle-income countries. AI-powered screening tools — trained on annotated retinal fundus photographs and OCT scans — offer a scalable solution, but they require large, precisely annotated datasets that few organisations can produce in-house.
How SCILabel Serves This Industry
Data Collection
We source retinal fundus photographs, optical coherence tomography (OCT) B-scans, and fundus fluorescein angiography datasets from ophthalmology clinics, eye hospitals, and public health screening programmes. All images are de-identified and supplied with consent documentation.
Data Annotation & Labeling
Our ophthalmology-trained annotators label retinal fundus images for diabetic retinopathy severity grading (ICDR scale), hard exudates, cotton wool spots, haemorrhages, neovascularisation, optic disc cupping (cup-to-disc ratio), and drusen. OCT B-scan annotation includes fluid segmentation, retinal layer delineation, and geographic atrophy mapping.
Data & Model Evaluation
Clinical evaluators benchmark model performance against ophthalmologist ground truth using AUROC, sensitivity at fixed specificity, and clinical grading concordance. We test for demographic bias across age, sex, and ethnicity subgroups.
Annotation Types & Formats
- Diabetic retinopathy grading (ICDR 0–4 scale) on fundus photographs
- Lesion-level annotation: haemorrhages, hard exudates, cotton wool spots, new vessels
- Optic disc and optic cup segmentation for glaucoma models
- OCT fluid segmentation: SRF, IRF, PED
- Retinal layer segmentation on OCT B-scans
Specialist Workforce Tracks
Track 2 (Medical Imaging): with ophthalmology specialisation: Optometrists, Ophthalmic Clinical Officers, Ophthalmologists.