Purpose-built for
AMD · DR · Glaucoma · Cardiovascular.
Four diseases account for over three-quarters of preventable adult blindness and a large share of undetected cardiovascular mortality worldwide. OCTON's dual-brain AI runs a dedicated deep-scan pipeline for each — pixel-level structural analysis, evidence-graded staging, and a closed-loop referral to a sub-specialty centre if it matters.
Age-related Macular Degeneration
OCTON's AMD pipeline detects the earliest structural changes at the macula — drusen morphology, retinal pigment epithelial changes, and subretinal fluid — and stages severity against the AREDS-2 classification. If it sees anything suspicious, you get a follow-up window, home-monitoring guidance (Amsler grid, high-dose AREDS supplementation triggers), and an instant referral packet to a macula specialist.
Staging follows Ferris et al., 2013 (Ophthalmology 120) and the AREDS-2 report. Deep-scan classification is corroborated on-device against reference drusen morphology from the IEEE Kaggle iChallenge-AMD dataset.
- Macula centred (fovea in mid-frame)
- Dim room · dilated pupil ideal (not required)
- Steady phone, 4–6 in / 10–15 cm distance
- Optional: pair a handheld fundus camera (Devices)
- Peng et al., Nature Communications 2019 — deep learning matches AREDS graders.
- Burlina et al., JAMA Ophthalmol 2017 — automated fundus grading.
- Grassmann et al., Ophthalmology 2018 — 13-class AMD DNN.
- Optic disc + macula visible (a 45° field is ideal)
- Two fields per eye (posterior pole + nasal)
- Higher-res via clinic camera boosts sensitivity for microaneurysms
- DICOM export from tabletop devices supported (Devices)
- Gulshan et al., JAMA 2016 — Google Health DR algorithm.
- Ting et al., JAMA 2017 — SELENA+, 71,896 images.
- Abràmoff et al., npj Digital Medicine 2018 — IDx-DR autonomous AI (FDA-cleared).
Diabetic Retinopathy
For anyone with type 1 or type 2 diabetes, OCTON's DR pipeline classifies severity against the ETDRS scale — from no retinopathy through mild, moderate, severe non-proliferative, and proliferative disease. We flag microaneurysms, dot-blot haemorrhages, cotton-wool spots, hard exudates, IRMA, and venous beading; then trigger the correct urgency of follow-up.
Grading follows the Early Treatment Diabetic Retinopathy Study (ETDRS) and the International Clinical Diabetic Retinopathy severity scale.
Catching the silent thief of sight.
Because glaucoma steals peripheral vision silently, patients rarely notice damage until 40% of optic nerve fibres are already lost. OCTON's deep-scan reads the optic disc, neuroretinal rim and RNFL quantitatively — spotting the pattern before symptoms appear.Ref: Tham Y-C et al., Ophthalmology 2014; Bourne R et al., Lancet Global Health 2016.
The pipeline follows the ISNT rule for neuroretinal rim thickness, quantifies vertical cup-to-disc (C:D) ratio, checks OD/OS asymmetry, and looks for disc haemorrhages and RNFL wedge defects. Whenever OCTON sees a red flag, we push an urgent recommendation for tonometry + visual-field testing.
Staging follows the European Glaucoma Society Terminology & Guidelines (EGS-5, 2020) and the ISGEO consensus. Automated OCTON grading validated against reference datasets from the REFUGE-2 and OIA-ODIR-5K challenges.
- Optic disc centred, not the macula
- Bilateral capture — OCTON compares OD vs OS asymmetry
- Undilated is fine if the disc edges are sharp
- OCT-quality boost from any Bluetooth camera (Devices)
- Ting DSW et al., npj Digital Medicine 2019 — deep learning for glaucoma detection, AUC 0.94.
- Li Z et al., Ophthalmology 2018 — 48,000-image glaucomatous optic-disc classifier.
- Liu H et al., JAMA Ophthalmol 2019 — external validation across 26 clinics.
- Include optic disc + macula so the vascular arcades are visible
- Sharp focus on arterioles + venules (thickness matters)
- Two eyes ideally — cross-check AV-ratio symmetry
- Provide age + BP + smoking status in Profile for calibrated risk
- Poplin R et al., Nature Biomedical Engineering 2018 — age, sex, smoking, BP and cardiac events predicted from fundus alone.
- Rim TH et al., Circulation 2020 — retinal-based coronary-artery-calcium prediction.
- Wong TY et al., Ophthalmology 2001 — Atherosclerosis Risk in Communities (ARIC) retinal vessel study.
- Cheung CY et al., Lancet Digital Health 2022 — dementia + stroke risk from retinal AI.
The retina is a window on your arteries.
Retinal micro-vessels mirror the systemic circulation exactly. OCTON reads arteriovenous ratio, arteriolar narrowing, AV nicking, cotton-wool spots and hard exudates — then maps them to hypertensive-retinopathy staging and a Poplin-style 10-year major adverse cardiac event (MACE) risk score. When the retina looks bad long before the heart does, we get you to your GP with hard evidence.
Grading follows the classic Keith–Wagener–Barker classification of hypertensive retinopathy, plus a MACE risk score derived from retinal biomarkers per Poplin et al. 2018. Estimates are age- and sex-adjusted using the Framingham / PREDICT baseline.
Deeper scans with a paired camera.
Pair a Bluetooth-capable handheld fundus camera (Optomed Aurora IQ, Remidio FOP), a passive optical clip-on (D-EYE, PEEK Retina, iExaminer 5.1) or a tabletop DICOM camera (Zeiss VISUCAM). OCTON accepts JPEG, PNG, HEIC and full DICOM — the deep-scan pipeline runs the same way on every source.
Ready for a Deep Scan?
One capture. Two disease pipelines. Instant severity staging, evidence citations, and — if it matters — a same-day referral packet to a sub-specialty centre near you.