Screen for diabetic retinopathy, glaucoma, and hypertensive retinopathy from a single fundus photograph. Designed for primary care clinics with non-specialist staff.
Get started View evidenceICDR Grade 0-4 classification. Detects microaneurysms, haemorrhages, exudates, and neovascularisation from standard fundus images.
Cup-to-disc ratio estimation with suspect flagging. Identifies optic disc pallor, rim thinning, and RNFL defect proxies.
Wong-Mitchell grading (Grade 0-4). Detects arteriolar narrowing, AV nicking, flame haemorrhages, and papilloedema.
| Disease | AUC | Sensitivity | Specificity | Classification |
|---|---|---|---|---|
| Diabetic Retinopathy | 98.5% | 95.3% | 94.1% | ICDR Grade 0-4 |
| Glaucoma | 97.1% | 86.2% | 99.1% | CDR + RNFL Proxy |
| Hypertensive Retinopathy | 99.1% | 88.9% | 99.9% | Wong-Mitchell |
| Macro Average | 98.2% | 90.1% | 97.7% | 3-disease combined |
EfficientNet-V2-L trained on 13,434 fundus images (ODIR-2019 + Glaucoma Detection). Validated on held-out test set.
Take a fundus photograph using any compatible camera. Non-mydriatic, no eye drops required. Takes 30 seconds per eye.
Upload the image to the platform. AI analyses for all three diseases simultaneously and returns results in under one second.
Ophthalmologist validates AI findings remotely. Confirmed cases are referred to the nearest tertiary centre with a complete clinical report.
Principal Investigator
Universiti Malaya
Consultant Ophthalmologist
Collaborator
AI and Technology Lead
Reporting Contact
General Physician, CRC
Contact for Contracting
Clinical Advisor
Contact for Contracting